2:33 p.m.
http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/
What is your experience ?
Cheers,
Valent.
--
http://kernelreloaded.blog385.com/
linux, blog, anime, spirituality, windsurf, wireless
registered as user #367004 with the Linux Counter, http://counter.li.org.
ICQ: 2125241, Skype: valent.turkovic
2:53 p.m.
On Mon, Mar 24, 2008 at 2:33 PM, Valent Turkovic
<valent.turkovic(a)gmail.com> wrote:
http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/
What is your experience ?
The general wisdom from fedora power users is:
- Fedora for the desktop / workstation
- CentOS for the home server or other unsupported server
- RedHat for a serious server or otherwise when support is needed
* by support I mean paid support
Also, SELinux isn't nearly that hard. I wouldn't put a box on the open
internet without SELinux.
--
Fedora 7 : sipping some of that moonshine
( www.pembo13.com )
4:42 p.m.
Albert Graham wrote:
I have installed hundreds of servers using Fedora and I have to say I've
had very few problems, kernel issues are not really the fault of the
Fedora team, sometimes you hit quirks but these do get sorted out.
How many years have you maintained these servers and how much damage
does downtime cause?
--
Les Mikesell
lesmikesell(a)gmail.com
11:48 a.m.
Valent Turkovic wrote:
On Mon, Mar 24, 2008 at 10:01 PM, Albert Graham
<agraham(a)g-b.net> wrote:
> Valent,
>
> You get out of it what you put in it, this guy clearly could not be
> bothered to look into issues that he was having or why things had
> changed - which is called progress.
>
> I have installed hundreds of servers using Fedora and I have to say I've
Why do you use Fedora and not CentOS for servers?
What is the benefit for you using Fedora for servers? How often do you
upgrade to latest version of fedora those hundereds of servers you
installed?
Just to add another datapoint. I do use Fedora for servers - 6 of them
doing a variety of jobs, as well as for workstations.
All of our servers are running the latest version of Fedora, and are
usually updated a couple of weeks after a new version comes out. We do
see some minor problems occasionally when updating, but we've never had
to back out of an update, and the fix for all problems has always been
simple.
Whilst these servers may not count as 'mission critical' by some of your
standards I would get a lot of disgruntled users if they were to
disappear for more than a day or so.
Having done upgrades (not reinstalls) on these machines since FC1 (for
the oldest one) I've never yet had any downtime on any service due to
problems from Fedora which required anything more than 5 minutes work to
restart a service.
Personally I find the rolling updates an easy way to manage these
machines. The pain of updating comes in small, easily managed chunks
with none of the major headaches you can get when doing a major update.
I reckon on losing about a days worth of uptime across all machines (ie
a couple of hours each) every 6 months to do the update. I'm sure that
people with CentOS installs allow this much time for scheduled
maintainance anyhow so I'm not really behind much on that front.
It's true that bits of Fedora blow up now and then, but if you're on
common hardware (our servers are all Dell PowerEdge machines) anything
major gets fixed very quickly. Also, most of the things which break (X,
sound, desktop applications etc) don't normally matter to a server.
PluseAudio may be broken on my servers right now, but I don't really
care. As has been mentioned, most server software is very stable and
doesn't often break these days.
I appreciate the ease of only having to deal with one platform and
always having the latest version of everything I want to work with, and
the downsides of that have never caused me any significant inconvenience.
Using Fedora for a server may not be as daft as some people think.
Simon.
6:43 p.m.
David G. Mackay wrote:
On Tue, 2008-03-25 at 10:07 -0500, Les Mikesell wrote:
> There was a point in mid-FC5 where an update kernel would not boot on an
> IBM Xseries 225 (fairly mainstream dual-xeon boxes, I thought) without
> doing a motherboard bios update - after which the older kernels wouldn't
> boot. And of course there were no warnings about this or much of a
> reason to expect it to work after doing the bios flash either. And
> there have been several updates that failed to boot on popular Dell and
> IBM MPT scsi controllers - even one late into FC6 which was otherwise
> pretty stable.
I'm not arguing that it doesn't happen. I've experienced a few. The
original post was directed to a fellow who wants to run a home server on
The original post was by valent.turkovic and actually referred to a
document written by someone who'd decided Fedora was a poor choice for a
server, and invited comment.
his desktop. And, I haven't seen anything yet that would cause
more
than a few minutes downtime while switching to the backup partition
after noticing the problem.
Now, if we're talking a system that's handling financial transactions
that has to be available 24/7, then it's another situation entirely.
I've been in shops like that, and am aware of the practices necessary to
sustain that. I don't, however, need a disaster recovery plan with a
hot backup site for my video and music collection.
I would think that a home Linux user running a server controlling his
Internet connexion and maybe providing mail and http caching is fairly
common. While maybe not life-threatening, its sudden lack might be
pretty inconvenient.
_Mine_ is running CentOS4 and is on a UPS (the UPS because used good
ones can be bought pretty cheaply at auction).
--
Cheers
John
-- spambait
1aaaaaaa(a)coco.merseine.nu Z1aaaaaaa(a)coco.merseine.nu
-- Advice
http://webfoot.com/advice/email.top.php
http://www.catb.org/~esr/faqs/smart-questions.html
http://support.microsoft.com/kb/555375
You cannot reply off-list:-)
7:49 p.m.
Les Mikesell wrote:
Roger Heflin wrote:
>
>> I can't recall ever being in a position of "having to bring in new
>> hardware". What scenario forces this issue on you? I haven't
>> noticed a shortage of vendors who will sell RHEL supported boxes.
>> But it sounds like you have an interesting job...
>>
>
> More cpu power needed to do the job. And the new boxes aren't
> officially RHEL supported (and sometimes won't even boot with the
> latest update-but will work with the latest fedora/kernel.org).
Something faster than IBM could sell you?
At the time, yes, this was before IBM sold AMD stuff, and the early though
troublesome Athlon's were faster than the Intel stuff.
> I had a subset of machines (about 250 machines) all of which had
> reached about 500+ days of uptime (the uptime counter rolled over)
Wasn't that fixed circa RH8? I had some 7.3 machines roll over twice.
It was pre-RH8.
> The issue with all OSes is that no one tests enough to catch these
> high MTBF issues, and in a big environment a machine crashing 1x per
> every 1000 days of uptime, comes to 1 machine a day crashing because
> of software, and typically the enterprise OSes aren't even close to
> that level, and while fedora is worse, it is just not that much worse.
I don't think RH7.3 with its final updates or Centos3.x (where x>1) had
anything approaching a software crash per 1000 days - at least not in
the base system and common services. I mostly skipped the 4.x series
because I didn't trust the early 2.6 kernels at all, but 5.1 seems solid.
Both of them have issues if you are running NFS servers with lots of clients,
other than that they are pretty stable, but if you are relying on NFS heavily
that is a show-stopper, but once you get a working stable setup if you really
want stability you don't touch it, no matter how good anyone tests things, they
will miss something, and things are worse the more different applications you
are running, all doing different odd things each of which may find one the bugs
no one at Redhat/Suse found in their testing.
And on top of that I have had trivial driver changes in the enterprise OSes
cause huge performance regressions (an FC driver update changed the queue depth
to 64-which caused the speed to be 30% of what it was before on certain external
FC raid disk arrays-this affected SLES9sp3 (9sp[12] kernel was ok), SLES10, any
kernel.org with the newer driver, RHEL4(all of them at the time)), so no update
can be counted on to not cause issues, this error was not seen by the driver
maintainer until they got one of the external arrays to test with and saw it
compared to a competitors board that was 3x faster under the newer kernel, but
almost identical under the older kernel, and both RHEL and Sles testing did not
catch it, to fix it we actually had to update to an unreleased driver that
allowed the queue depth to be changed down (none of the updates at the time
fixed it), and wait for a update on Sles. To get this fixed it was far easier
to work with the upstream driver maintainer and get them to push the update to
the enterprise vendors than to try to get the enterprise vendors to find and fix
the problem. I was told by a different upstream maintainer that typically the
enterprise vendors pushed any serious issues directly to them, and did very
little with the issues themselves, if you could get it past their first line
support people.
The big problem is that the testing has to include it does not crash, it runs
roughly the same speed as before, and it still gives the same answer, and even
if one runs every test they know about, some configuration will still get
through for a given setup.
I guess my experience is that even with Enterprise updates at least 25-50% of
the time there is a serious regression (speed, crash, wrong answer), one has to
carefully consider what do I gain by doing that update. So the testing require
for a full update is really no better than the testing required to go from F7 to
F8, and fedora updates puts out new kernels faster so getting a fix into the
stream is a lot faster than with the enterprise oses, and once you get one that
works on a given piece of HW correctly you stop updating, some of the things I
have ran into on an update are things you would have never though to test for,
so you have to watch out on any update-and it is just best to only update when
required to.
Some of the customers I used to support typically stayed on what was shipped
with the machine, because their validation procedures were fairly extensive, and
the update not worth it when no useful features were to be gained.
9:46 p.m.
On Wed, 2008-03-26 at 09:41 +1000, Da Rock wrote:
I'm going to put my 2c in and say if you want a server that is
stable,
doesn't need constant updating, can be changed without rebooting (if
you're clever), and can get easy, continuous support throughout the
ages, then go FreeBSD and be done with it.
The lists are informative and will help no matter what, and I've run a
web server, dns, and more on a little shitbox which won't run anything
else for 27 days without rebooting and making fairly major changes in
the process. Its only a pII with a 3Gb drive.
I agree. I've had several knowledgeable folks suggest it, but I've just
not gotten around to it. How are they about security fixes?
I believe you use the right tools for the right job- Fedora CAN be
used
as a server, but you have to ask: should you? FreeBSD is primarily
designed as a server, and most ISP's and other services use it for its
reliability. I know of a server that was up for over 3 years without
stopping running FreeBSD. I believe it's also possible to hot swap
kernels on FreeBSD- something I wouldn't try with linux.
I don't think hot swapping kernels is anywhere in the near future on
linux. (But I'd be happy to be corrected on this) It does bring to
mind a recent situation. A couple of years ago, I was involved in a
project with a client. As part of it, I set up a server running Apache,
Cyrus Imap, Squirrelmail, and Samba. The project reached completion,
and the server got taken over by another group. I got a call about two
months ago. It seems that the new group had been using the server
totally unmodified, and without technical support. Web access had
stopped working, but email was fine, and other folks could access the
web pages being served by Apache. After over two years of running
unattended, the Squid log had reached the 2GB file size limit, so Squid
stopped running. Once I cleared that, Squid came back up, and they were
happy again. When the box can run over two years without ANY technical
support, it's hard to convince them that they need a support contract.
Linux is great for desktop apps and can be used for server
solutions,
but it was designed for desktop use- so thats its strong point. That
said FreeBSD or any other Unix system is a lot harder to setup as a
desktop system, but it can be done too.
Well, I've had my rant... I feel happier now. I'm off...
I dunno about you, but I'm on this list mostly for the things that I can
learn here. Thanks for reminding me about FreeBSD.
Dave
9:23 p.m.
On Wed, 2008-03-26 at 08:36 +0900, John Summerfield wrote:
Not Fedora exactly (it's the f9a), but yum has conspired with rpm
against me to remove every working kernel. yum ignored how many kernels
I wanted to keep, and (apparently) upgraded (as opposed to installed)
the kernel.
It _could_ happen in Fedora, to anyone who had just one kernel (eg a new
install).
Yes, and that's why I use a VM to install and test the new releases. If
it completely fries the VM, I'm only out a few minutes time.
> Having a VM running a server in this environment may not be the
absolute
> best practice, but it's certainly feasible.
That's not the point I was speaking about. Running a VM _under Fedora_
is. A VM under CentOS or RHEL is altogether different. RHEL/CentOS is
less likely to break with a new update, and since it presents
conservative "hardware" to the guest, Fedora is less likely to break too.
What can I say? I do have multiple machines, and have dedicated
servers, and I want something with a long lifecycle for those servers
doing fairly standard things, like dns, http, imap, samba, etc. At the
same time, I like to keep tabs on fairly state-of-the-art developments
so that I can get a good idea about where my own development efforts
should go. I have the luxury of multiple machines. If I didn't,
running a VM with Fedora on the desktop and Centos as the guest would be
a likely choice. I have some hardware (a tv capture card, for one) that
isn't implemented in virtualized hardware. FWIW, I've had pretty good
success running QEMU-KVM out of rawhide with multiple guest OSes. If I
relied on the version available in RHEL or Centos, I wouldn't be nearly
as happy. In the case of QEMU/KVM, the bleeding edge stuff is working
better. Of course, that's just my experience, and I guess that it's
possible that I've been fantastically lucky, but I don't think so.
Dave
8:34 a.m.
On Sat, 2008-04-12 at 21:32 +0700, Strong wrote:
On Mon, 24 Mar 2008 22:24:54 -0500 "David G. Mackay"
<mackay_d(a)bellsouth.net> wrote:
> If you don't have
> hardware assisted virtualization, you can still run vmware, qemu, xen,
How I can find out if my hardware virtualization capable?
grep vmx /proc/cpuinfo
poc
3:21 p.m.
Les Mikesell wrote, On 03/25/2008 03:23 PM:
David G. Mackay wrote:
>
> Now, if we're talking a system that's handling financial transactions
> that has to be available 24/7, then it's another situation entirely.
I assumed that the word 'servers' in the subject meant that the services
mattered to someone...
for many of us, server = a machine running at least one of the following, NFS,
SAMBA, bind, yp, LDAP, Apache....
> I've been in shops like that, and am aware of the practices
necessary to
> sustain that. I don't, however, need a disaster recovery plan with a
> hot backup site for my video and music collection.
If reliability over some long interval isn't important to you, then this
thread probably isn't very interesting.
I would say reliability over a long interval is not the problem here, as I
have had some VERY reliable FC servers (handling 30+ folks.) for years[0].
Any admin who updates a critical server without doing some testing on a
similar machine, should no longer be allowed to do the job, even when those
updates are paid for from RH or MS. Things like 'we know kernel 2.6.24.x-fc8
broke on ..." is only another chunk of info for the admin, [s]he should be
watching for the same things if using a Enterprise editions, because they
sometimes[1] have those problems too[2].
However, using FC means that that 'long interval' has to be ended[3] sometime
between the time that the vendor drops doing security updates for that distro,
and when the admin can no longer keep the box patched up to the appropriate
level on the budget available. Fedora makes the 'long interval' 1year+, and
Enterprise editions make it ~7years[4]. THIS is why fedora should loose to
RHEL|CentOS when dealing with a server that needs to be in service for much
longer than a year.
[0] I miss fedora legacy.
[1] Granted, they USUALLY have the problems less OFTEN, severity is often the
same.
[2] read the autofs, DRBD and Linux-HA lists for a while.
[3] ended by upgrade to a new distro version or retirement of the machine.
[4] http://www.redhat.com/security/updates/errata/
--
Todd Denniston
Crane Division, Naval Surface Warfare Center (NSWC Crane)
Harnessing the Power of Technology for the Warfighter
7:34 a.m.
Tomasz Torcz wrote:
>> > And the second you add a driver and/or XFS on to RHEL5
you are
>> > now tainted and *UNSUPPORTED*.
>> Compared to Fedora where you are *UNSUPPORTED* at the offset?
> is XFS even in the vanilla kernel?
It is in vanilla 2.4 since December 2003. It is in current kernel
since september 2002 (kernel 2.5.36).
Doesn't it still have problems with 4k stacks on 32 bit machines?
--
Les Mikesell
lesmikesell(a)gmail.com
4:58 a.m.
> > And the second you add a driver and/or XFS on to RHEL5 you
are
> > now tainted and *UNSUPPORTED*.
> Compared to Fedora where you are *UNSUPPORTED* at the offset?
is XFS even in the vanilla kernel?
It is in vanilla 2.4 since December 2003. It is in current kernel
since september 2002 (kernel 2.5.36).
--
Tomasz Torcz
1:58 a.m.
On Tue, Mar 25, 2008 at 8:48 PM, David G. Mackay <mackay_d(a)bellsouth.net> wrote:
On Tue, 2008-03-25 at 18:41 -0500, Les Mikesell wrote:
> > Fedora seems to be getting away from this, but one of the selling points
> > for Slackware used to be that you could take some of the older hardware
> > that wouldn't be able to handle the latest Microsoft OS, and still do
> > useful things with it under Linux.
>
> Centos 3.x is still a moderately good choice for this. In fedora terms
> it would be very, very similar to FC1 (2.4 kernel, etc.) but still
> getting updates.
I was thinking of updating my firewall/router box to Centos 4.6. Do you
have any thoughts/warnings on this?
Dave
Why not Centos 5?
--
Fedora 7 : sipping some of that moonshine
( www.pembo13.com )
11:14 p.m.
David G. Mackay wrote:
>> Fedora seems to be getting away from this, but one of the selling points
>> for Slackware used to be that you could take some of the older hardware
>> that wouldn't be able to handle the latest Microsoft OS, and still do
>> useful things with it under Linux.
> Centos 3.x is still a moderately good choice for this. In fedora terms
> it would be very, very similar to FC1 (2.4 kernel, etc.) but still
> getting updates.
I was thinking of updating my firewall/router box to Centos 4.6. Do you
have any thoughts/warnings on this?
It will work, but I don't see any compelling reason to choose it now.
5.x is newer/better and still old enough not to expect any surprises.
3.x is smaller, less bloated but still supported with updates. If you
run any other services on your gateway box (mail/web/ftp/file shares,
etc.) you might like SME server (http://www.contribs.org), though. I
think it is still based on Centos 4.x code - but managed with a simple
web interface.
--
Les Mikesell
lesmikesell(a)gmail.com
9:56 p.m.
On Wed, 2008-03-26 at 08:43 +0900, John Summerfield wrote:
> I'm not arguing that it doesn't happen. I've
experienced a few. The
> original post was directed to a fellow who wants to run a home server on
The original post was by valent.turkovic and actually referred to a
document written by someone who'd decided Fedora was a poor choice for a
server, and invited comment.
Sorry, let me rephrase that. The post to which I originally responded
was from a fellow who wanted to run a home server on his desktop.
I would think that a home Linux user running a server controlling his
Internet connexion and maybe providing mail and http caching is fairly
common. While maybe not life-threatening, its sudden lack might be
pretty inconvenient.
Depending on the mood that my wife is in when her internet connection
gets severed, it could indeed be life threatening.
_Mine_ is running CentOS4 and is on a UPS (the UPS because used good
ones can be bought pretty cheaply at auction).
Mine are on UPSes as well (new ones aren't that expensive either).
Still, the wireless access point is in a different part of the house,
and isn't on a UPS. Fortunately, the wife gets vexed with the power
company on the occasions where that's a problem.
Dave
9:48 p.m.
On Tue, 2008-03-25 at 18:41 -0500, Les Mikesell wrote:
> Fedora seems to be getting away from this, but one of the
selling points
> for Slackware used to be that you could take some of the older hardware
> that wouldn't be able to handle the latest Microsoft OS, and still do
> useful things with it under Linux.
Centos 3.x is still a moderately good choice for this. In fedora terms
it would be very, very similar to FC1 (2.4 kernel, etc.) but still
getting updates.
I was thinking of updating my firewall/router box to Centos 4.6. Do you
have any thoughts/warnings on this?
Dave
8:25 p.m.
Les Mikesell wrote:
> I've never been "forced to re-install some wildly
different version".
The FC1->2, FC3-4, FC6-7 transitions had huge differences.
I wouldn't call them _huge_. Large, yes.
But they didn't force me to _re_ install anything.
As it happens, I always keep the previous version of Fedora on my server,
so I wouldn't need to re-install even if I wanted to revert.
But in fact I have never needed to revert.
Sound has been hit-or-miss with huge differences in each version.
In my experience, sound was always bad in Fedora, is bad today,
and I fear always will be bad.
--
Timothy Murphy
e-mail (<80k only): tim /at/ birdsnest.maths.tcd.ie
tel: +353-86-2336090, +353-1-2842366
s-mail: School of Mathematics, Trinity College, Dublin 2, Ireland
7:25 p.m.
On Tue, Mar 25, 2008 at 8:04 PM, John Summerfield
<debian(a)herakles.homelinux.org> wrote:
You seem to have a fair appreciation of the key points. Unfortunately, a
lot of folk get "Fedora Unleashed" - often for an old or even obsolete
release - and have at it without understanding those key points. The
author of the original article seems to have been one such.
I'm curious, just for clarification, what key points did the author of
the original article not understand?
-Mauriat
7:04 p.m.
max bianco wrote:
On Tue, Mar 25, 2008 at 2:31 PM, Les Mikesell
<lesmikesell(a)gmail.com> wrote:
> max bianco wrote:
>>
>> am I the only person that voluntarily uses Fedora? I have only been
>> using it a short time but i understood it was a moving target when i
>> signed on. This thread reminds me of KL. It wasn't that long ago so i
>> am sure most of you remember to whom i am referring.
>> Fedora is what it is and some have more experience than others. Is there
>> some other lesson to take away from this thread?
> Fedora tries to stay very close to current software development. That
> means you get the new features as well as the new bugs. On a desktop or
> as a second, non-critical machine that is a reasonable tradeoff,
> especially if you use the applications where the new features matter.
> This thread is mostly about servers where downtime is to be avoided at
> (almost) any cost and you want things to keep working for years. It
> isn't about fedora being inherently bad, it just isn't suitable for some
> uses.
>
>
I know it isn't suitable for some uses. I tried to learn what I could
about Fedora before I actually installed it on my machine. It seems
however that some people have unrealistic expectations. The bleeding
edge and stable are not words i would expect to fit into the same
sentence under any circumstances. Having said that it has been very stable
but you just did:-)
for me, and by stable i mean that it has never failed to boot, have i
had
issues? Sure but the basic functionality has always been there for me. Maybe
some features break from time to time but i have always been able to access
the web to solve my problems, would i use it for a server? maybe but I'd not
have much room to complain if I tried to ride the bleeding edge and fell off
the cliff because of it.
You seem to have a fair appreciation of the key points. Unfortunately, a
lot of folk get "Fedora Unleashed" - often for an old or even obsolete
release - and have at it without understanding those key points. The
author of the original article seems to have been one such.
--
Cheers
John
-- spambait
1aaaaaaa(a)coco.merseine.nu Z1aaaaaaa(a)coco.merseine.nu
-- Advice
http://webfoot.com/advice/email.top.php
http://www.catb.org/~esr/faqs/smart-questions.html
http://support.microsoft.com/kb/555375
You cannot reply off-list:-)
6:58 p.m.
Les Mikesell wrote:
John Summerfield wrote:
> Arthur Pemberton wrote:
>> On Mon, Mar 24, 2008 at 2:55 PM, Tim Alberts <talberts(a)msiscales.com>
>> wrote:
>>> Honestly I feel like a backstabber by using CentOS because I've been
>>> with Fedora since before Fedora (RedHat 5.1 was my first Linux
>>> experience), and using CentOS is reaping the benefits with no
>>> contribution.
>>
>> You realize that technically speaking Redhat|Centos are infact Fedora,
>> albeit mature versions of it.
>>
>
> Not really. The testing that goes into RHEL is quite different.
And yet there are packages in RHEL5 that weren't rebuilt from their FC6
versions...
I'm not sure that's so[1], but if it is, so what?
It's the _distro_ testing that is the key difference. RH is betting its
future on its RHEL releases. Problems with Fedora are trivially
important in comparison (and arguably not RH's problems at all).
[1] I've seen EL packages that seem to be FC packages, but that might
mean no more than the same spec file and that the spec file is not
sensitive to the build environment. THis, for example, will look like a
RHEL package even if built on Fedora:
Name: fedora-ds
Version: 1.0
Release: 2.RHEL4
while some others expand %{dist} like so:
Release: 2%{?dist}
--
Cheers
John
-- spambait
1aaaaaaa(a)coco.merseine.nu Z1aaaaaaa(a)coco.merseine.nu
-- Advice
http://webfoot.com/advice/email.top.php
http://www.catb.org/~esr/faqs/smart-questions.html
http://support.microsoft.com/kb/555375
You cannot reply off-list:-)
6:41 p.m.
David G. Mackay wrote:
On Wed, 2008-03-26 at 07:25 +1030, Tim wrote:
> One might think that if you're using a personal computer as a server,
> that by the time a several-year lifespan OS reached end-of-life, you'd
> probably want to be taking advantage of new hardware, as well. So an
> upgrade by building a new machine, copying data over to it, and swapping
> it over for the old server, would seem the prudent way to go about it.
Fedora seems to be getting away from this, but one of the selling points
for Slackware used to be that you could take some of the older hardware
that wouldn't be able to handle the latest Microsoft OS, and still do
useful things with it under Linux.
Centos 3.x is still a moderately good choice for this. In fedora terms
it would be very, very similar to FC1 (2.4 kernel, etc.) but still
getting updates.
--
Les Mikesell
lesmikesell(a)gmail.com
6:41 p.m.
On Tue, 2008-03-25 at 18:10 -0500, David G. Mackay wrote:
On Tue, 2008-03-25 at 14:23 -0500, Les Mikesell wrote:
> David G. Mackay wrote:
> >
> > Now, if we're talking a system that's handling financial transactions
> > that has to be available 24/7, then it's another situation entirely.
>
> I assumed that the word 'servers' in the subject meant that the services
> mattered to someone...
To quote the folks from (now defunct) Data General, "Not everything
worth doing is worth doing well." The amount of concern for a web
server that's primarily serving up the family photos to the remote
relatives "matters", but not necessarily much. The services in F9 Beta
"matter", but I'm setting up my test environment in a VM, so if it goes
KFB then I haven't lost anything much.
> > I've been in shops like that, and am aware of the practices necessary to
> > sustain that. I don't, however, need a disaster recovery plan with a
> > hot backup site for my video and music collection.
>
> If reliability over some long interval isn't important to you, then this
> thread probably isn't very interesting.
You still don't see the cost/benefit analysis that I'm applying to a
home server. I can tolerate some downtime, but it's nice to have a
distro with a longer life cycle so that I don't have to do major
upgrades to my server every six to nine months just to keep it up to
date. My primary consideration there would be security updates, and
yes, I realize that I could skip releases with Fedora without losing the
security updates. By the same token, a problem in F7 isn't likely to
get the same level of response as one in F8.
Dave
I'm going to put my 2c in and say if you want a server that is stable,
doesn't need constant updating, can be changed without rebooting (if
you're clever), and can get easy, continuous support throughout the
ages, then go FreeBSD and be done with it.
The lists are informative and will help no matter what, and I've run a
web server, dns, and more on a little shitbox which won't run anything
else for 27 days without rebooting and making fairly major changes in
the process. Its only a pII with a 3Gb drive.
I'll eventually use it as a squid box and maybe mail server unless I get
new hardware.
I believe you use the right tools for the right job- Fedora CAN be used
as a server, but you have to ask: should you? FreeBSD is primarily
designed as a server, and most ISP's and other services use it for its
reliability. I know of a server that was up for over 3 years without
stopping running FreeBSD. I believe it's also possible to hot swap
kernels on FreeBSD- something I wouldn't try with linux.
Linux is great for desktop apps and can be used for server solutions,
but it was designed for desktop use- so thats its strong point. That
said FreeBSD or any other Unix system is a lot harder to setup as a
desktop system, but it can be done too.
Well, I've had my rant... I feel happier now. I'm off...
6:37 p.m.
Roger Heflin wrote:
> I can't recall ever being in a position of "having to bring in new
> hardware". What scenario forces this issue on you? I haven't noticed
> a shortage of vendors who will sell RHEL supported boxes. But it
> sounds like you have an interesting job...
>
More cpu power needed to do the job. And the new boxes aren't
officially RHEL supported (and sometimes won't even boot with the latest
update-but will work with the latest fedora/kernel.org).
Something faster than IBM could sell you?
I had a subset
of machines (about 250 machines) all of which had reached about 500+
days of uptime (the uptime counter rolled over)
Wasn't that fixed circa RH8? I had some 7.3 machines roll over twice.
The issue with all OSes is that no one tests enough to catch
these high MTBF issues, and in a big environment a machine crashing 1x
per every 1000 days of uptime, comes to 1 machine a day crashing because
of software, and typically the enterprise OSes aren't even close to that
level, and while fedora is worse, it is just not that much worse.
I don't think RH7.3 with its final updates or Centos3.x (where x>1) had
anything approaching a software crash per 1000 days - at least not in
the base system and common services. I mostly skipped the 4.x series
because I didn't trust the early 2.6 kernels at all, but 5.1 seems solid.
--
Les Mikesell
lesmikesell(a)gmail.com
6:36 p.m.
David G. Mackay wrote:
On Tue, 2008-03-25 at 20:51 +0900, John Summerfield wrote:
> There's losing and there's really losing. If Fedora won't boot, you have
> the services of neither until it's fixed.
>
> A little shore of Best Practice.
I don't think that anyone was talking about HA-Linux and redundant power
Nor was I.
supplies here, John. It's fairly trivial to set up a separate
partition
with a fallback OS. I routinely do this with Fedora on one of my
machines, and then rotate versions of Fedora between the two partitions
as the releases become available. If an update makes the newer version
unbootable (extremely rare that you can't just use grub to boot the
previous kernel), then there's always the old version to boot into.
Not Fedora exactly (it's the f9a), but yum has conspired with rpm
against me to remove every working kernel. yum ignored how many kernels
I wanted to keep, and (apparently) upgraded (as opposed to installed)
the kernel.
It _could_ happen in Fedora, to anyone who had just one kernel (eg a new
install).
Having a VM running a server in this environment may not be the absolute
best practice, but it's certainly feasible.
That's not the point I was speaking about. Running a VM _under Fedora_
is. A VM under CentOS or RHEL is altogether different. RHEL/CentOS is
less likely to break with a new update, and since it presents
conservative "hardware" to the guest, Fedora is less likely to break too.
--
Cheers
John
-- spambait
1aaaaaaa(a)coco.merseine.nu Z1aaaaaaa(a)coco.merseine.nu
-- Advice
http://webfoot.com/advice/email.top.php
http://www.catb.org/~esr/faqs/smart-questions.html
http://support.microsoft.com/kb/555375
You cannot reply off-list:-)
6:24 p.m.
On Wed, 2008-03-26 at 07:25 +1030, Tim wrote:
One might think that if you're using a personal computer as a
server,
that by the time a several-year lifespan OS reached end-of-life, you'd
probably want to be taking advantage of new hardware, as well. So an
upgrade by building a new machine, copying data over to it, and swapping
it over for the old server, would seem the prudent way to go about it.
Fedora seems to be getting away from this, but one of the selling points
for Slackware used to be that you could take some of the older hardware
that wouldn't be able to handle the latest Microsoft OS, and still do
useful things with it under Linux. I have an old system that I'm using
as a firewall/router/name server. I have an old laptop that I'm about
to install F9 Beta on to use as a carputer to "serve" music to my in
vehicle stereo.
Dave
6:10 p.m.
On Tue, 2008-03-25 at 14:23 -0500, Les Mikesell wrote:
David G. Mackay wrote:
>
> Now, if we're talking a system that's handling financial transactions
> that has to be available 24/7, then it's another situation entirely.
I assumed that the word 'servers' in the subject meant that the services
mattered to someone...
To quote the folks from (now defunct) Data General, "Not everything
worth doing is worth doing well." The amount of concern for a web
server that's primarily serving up the family photos to the remote
relatives "matters", but not necessarily much. The services in F9 Beta
"matter", but I'm setting up my test environment in a VM, so if it goes
KFB then I haven't lost anything much.
> I've been in shops like that, and am aware of the practices
necessary to
> sustain that. I don't, however, need a disaster recovery plan with a
> hot backup site for my video and music collection.
If reliability over some long interval isn't important to you, then this
thread probably isn't very interesting.
You still don't see the cost/benefit analysis that I'm applying to a
home server. I can tolerate some downtime, but it's nice to have a
distro with a longer life cycle so that I don't have to do major
upgrades to my server every six to nine months just to keep it up to
date. My primary consideration there would be security updates, and
yes, I realize that I could skip releases with Fedora without losing the
security updates. By the same token, a problem in F7 isn't likely to
get the same level of response as one in F8.
Dave
5:55 p.m.
Les Mikesell wrote:
Roger Heflin wrote:
>
> Yes, and typically to support anything recent you have too many
> add-ons on the enterprise OSes, if you are in a fast moving enterprise
> environment RHEL won't work.
Fast moving and enterprise are words you don't usually see together.
Don't you have to keep decades-old processes running?
> RHEL is probably quite good for any of the nice simple static
> enterprise environments, but most would argue there you should
> probably lock everything down so tight that few kernel
> updates/userspace are even required for anything, the problem is in an
> environment were you are constantly having to bring in new hardware
> that does not work on the older release, where you cannot wait 6
> months for RHEL to catch up.
I can't recall ever being in a position of "having to bring in new
hardware". What scenario forces this issue on you? I haven't noticed a
shortage of vendors who will sell RHEL supported boxes. But it sounds
like you have an interesting job...
More cpu power needed to do the job. And the new boxes aren't officially RHEL
supported (and sometimes won't even boot with the latest update-but will work
with the latest fedora/kernel.org). You typically bring in around a large
enough set of new machines at a time (usually this was 100-200 machines) any
only update the pieces required to support that new machine, and then you run
some test to validate that it gives the correct answers for various jobs. It
is really a money issue, the new machine is 2x the speed but not yet supported
by RHEL, so you would need 2x the number of old supported machines for 2x the
cost or more. Reliability was required (50 or so disk servers any one of which
would cause at least partial loss of access to data).
The problem is validating that the new hardware/OS (or old hardware/just an huge
update) is the same process, so you change only what you need to, and you are
better off starting with as new as possible and going from there. We were
typically only updating the kernel, change userspace was even more dangerous as
a bad library update could change answers, so unless we found a library bug that
could not be easily worked around, we did not update it.
Some of the compute customers don't apply updates, it is too risky to cause
downtime/wrong answers, they fix the issues that they find, and then every 1-2
years they update the older stuff to what is currently being used/proven on the
newest machines, the situation you have is you have sets of machines with
slightly different OS loads (which is kind of nasty), but the other choice is
that you update everything all of the time (and that is even worse-as too many
different types of HW must be revalidated to still give the correct answer).
You end up with old servers that may be running what was originally determined
to be stable by testing and they aren't touched. In this environment the
testing is just as bad with any OS as any other, you start with something like
F8 and then update/downgrade any parts that fail you to something that works,
and they you don't touch it for several years. I had a subset of machines
(about 250 machines) all of which had reached about 500+ days of uptime (the
uptime counter rolled over), the 20 or so machines out of that set that failed
to reach that uptime were all HW issues (usually disk failures), but some more
lethal failures that required the retiring of the specific machine, it was not
typically worth paying for fixing the nastier issues (usually MB failures) so
they were retired, it was often cheap to just buy a something new that was
validated and much faster. The issue with all OSes is that no one tests
enough to catch these high MTBF issues, and in a big environment a machine
crashing 1x per every 1000 days of uptime, comes to 1 machine a day crashing
because of software, and typically the enterprise OSes aren't even close to that
level, and while fedora is worse, it is just not that much worse.
Roger
4:41 p.m.
Roger Heflin wrote:
Yes, and typically to support anything recent you have too many add-ons
on the enterprise OSes, if you are in a fast moving enterprise
environment RHEL won't work.
Fast moving and enterprise are words you don't usually see together.
Don't you have to keep decades-old processes running?
RHEL is probably quite good for any of the nice simple static
enterprise
environments, but most would argue there you should probably lock
everything down so tight that few kernel updates/userspace are even
required for anything, the problem is in an environment were you are
constantly having to bring in new hardware that does not work on the
older release, where you cannot wait 6 months for RHEL to catch up.
I can't recall ever being in a position of "having to bring in new
hardware". What scenario forces this issue on you? I haven't noticed a
shortage of vendors who will sell RHEL supported boxes. But it sounds
like you have an interesting job...
--
Les Mikesell
lesmikesell(a)gmail.com
4:25 p.m.
Les Mikesell wrote:
Roger Heflin wrote:
>
>>>
>>> Fedora ships buggy kernels. As stated previously this is more
>>> upstream's fault,
>>
>> Beg your pardon, but just because someone writes broken code that
>> _does not_ force a distribution to ship it.
>>
>
> The problem is that the enterprise OS's ship buggy kernels too,
Yes, but they very, very rarely introduce new bugs in an update that
breaks a previously working system.
> I have found at least one bug in most of the recent enterprise kernels
> (RHEL3, RHEL4, RHEL5 and SLES9-never tested SLES10), and some of those
> bugs were very very ugly had at least one of them had long since been
> fixed upstream, and at least one of those bugs *NEVER* existed in the
> kernel.org kernels at all.
And the timeline between reporting the bug and getting the fix was...?
Wait 3 months for the next update, and even there you had to change a kernel
parameter (after the new update allow the parameter to even be adjusted), and
this was for a bug that could be reproduced by "dd if=/dev/zero of=filename
bs=1M", all it took was a machine with 32GB of ram or more.
> And lets take RHEL5 initial release, it did not support file systems
> larger then 8TB (ext3 only, no XFS), it did not support Areca and
> 3ware PCIe controllers even though those drivers had been out for 6+
> months at the time they shipped RHEL5, and those are most definitely
> enterprise boards. And the second you add a driver and/or XFS on to
> RHEL5 you are now tainted and *UNSUPPORTED*.
Do you have some point here - like a system that always supports
everything and never fails? I usually don't have as much of a problem
having to fiddle with a new machine where you obviously aren't already
relying on it or getting known-supported hardware when I don't have time
to fiddle as I do with updates breaking previously working things. But,
have you tried the Centosplus kernels that put xfs and the drivers that
rhel removes back?
Yes, and typically to support anything recent you have too many add-ons on the
enterprise OSes, if you are in a fast moving enterprise environment RHEL won't work.
RHEL is probably quite good for any of the nice simple static enterprise
environments, but most would argue there you should probably lock everything
down so tight that few kernel updates/userspace are even required for anything,
the problem is in an environment were you are constantly having to bring in new
hardware that does not work on the older release, where you cannot wait 6 months
for RHEL to catch up. If you are really worried about stability on anything you
have to carefully cherry pick just the security fixes, and the fixes for
problems that you know that you have, you don't risk adding anything that you
don't need, everyone that is bitching about the Fedora updates breaking things,
but applies *ALL* of the updates without checking to see if it adds any features
that they need or fixes any problems that they have is just asking for trouble,
whether on a desktop or a server. If you manage fedora this way you can run
it as a server successfully.
4:13 p.m.
Rahul Sundaram wrote:
Roger Heflin wrote:
> And lets take RHEL5 initial release, it did not support file systems
> larger then 8TB (ext3 only, no XFS), it did not support Areca and
> 3ware PCIe controllers even though those drivers had been out for 6+
> months at the time they shipped RHEL5, and those are most definitely
> enterprise boards. And the second you add a driver and/or XFS on to
> RHEL5 you are now tainted and *UNSUPPORTED*.
This is not quite the whole picture. If a third party kernel module is
loaded, that module is specifically unsupported assuming a ISV is not
involved. If a problem occurs that is not related to the module and you
can demonstrate that by reproducing the problem after unloading it,
support must be provided.
Rahul
Rahul,
Yes, only if you have a separate support contract, and even with those the
support provide is utter crap. Redhat (and Suse) don't seem to have an
escalation procedure that actually works for anything serious, at the customer
level you have to jump through more hoops that any of the kernel.org developers
require, mainly because you have to spend days or weeks attempting to get past
the stupid useless suggestions of the first line support people that have no
clue that kernel deadlocks/crashes are almost impossible to cause with userspace
mis-configurations, and keep suggesting new userspace related things to change
to correct the issue, it is very clear that they are trained not to attempt to
solve everything, and don't understand when to pass it up.
Roger
4:06 p.m.
Arthur Pemberton wrote:
On Tue, Mar 25, 2008 at 2:17 PM, Ian Chapman
<packages(a)amiga-hardware.com> wrote:
> Roger Heflin wrote:
> >
> > The problem is that the enterprise OS's ship buggy kernels too, I have
>
> Of course they do, I'd be surprised to hear if anyone has shipped a 100%
> bug free kernel, whether it be Linux, Solaris, AIX, Windows or whatever.
> The point though is that RHEL is less of a moving target, if you current
> setup works then an update to the kernel is less likely to break
> anything, than say Fedora which frequently ships new versions of the
> kernel.
One would hope that that these people complaining abut imperfect
kernels in RHEL are find these bugs on their test boxes _before_ they
deploy to production.
Not likely, one of the bugs I found happens with a single application, the test
setup was run it in a job of 100 machines together, and by morning one of the
machines in the job will deadlock, not something that can be found with a single
test box, but it was consistent, the customer could run in on 3 separate sets of
100 overnight, and all 3 would have a single machine that locked up overnight,
reboot the 3 and/or exclude the 3 "bad" machines and try again, and a completely
different 3 machines locked up.
> > And the second you add a driver and/or XFS on to RHEL5 you
are
> > now tainted and *UNSUPPORTED*.
> Compared to Fedora where you are *UNSUPPORTED* at the offset?
is XFS even in the vanilla kernel?
Pretty much with RHEL you are unsupported once you install and configure it, at
that point you only really have update support, I have actually called their
support several different times with real actual kernel problems and the support
that was provided only wasted everyones time.
Yes, it is, Sles ships it, and it is in kernel.org kernels. Since RHEL does not
have XFS that means that RHEL is limited to a max 8TB filesystem (this is the
actual ext3 limit on 64bit machines), which eliminated it from usage for several
enterprise customers.
RHEL5 shipped without drivers for Areca and 3ware PCI-e cards even though the
drivers were out for months at the time (the drivers were in vanilla 2.6.19, 1
rev over 2.6.18 used by RHEL5, but Redhat for some reason did not backport the
driver for what is basically an enterprise class raid board for the initial
release).
3:55 p.m.
On Tue, 2008-03-25 at 16:21 -0400, Todd Denniston wrote:
ended by upgrade to a new distro version or retirement of the
machine.
One might think that if you're using a personal computer as a server,
that by the time a several-year lifespan OS reached end-of-life, you'd
probably want to be taking advantage of new hardware, as well. So an
upgrade by building a new machine, copying data over to it, and swapping
it over for the old server, would seem the prudent way to go about it.
--
(This computer runs FC7, my others run FC4, FC5 & FC6, in case that's
important to the thread.)
Don't send private replies to my address, the mailbox is ignored.
I read messages from the public lists.
2:48 p.m.
On Mon, Mar 24, 2008 at 08:33:41PM +0100, Valent Turkovic wrote:
http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/
What is your experience ?
"Fedora is about the rapid progress of Free and Open Source software."
If that covers your needs for a server, that's OK. But usually most
people will associate a server with high availability and low
maintenance, e.g. less rapid progress and more conservative steps to
not endanger their setups.
So actually the answer is not that "Fedora makes a terible server",
but that the question about Fedora being run as a server needs to be
placed in Fedora's frame to be valid. And if you're up to upgrading
often and using sandbox systems to check whether the functionality you
needed is still mapped properly, then go for it.
But I very much disagree with the last sentence:
"However I will have more realistic expectations from software
considered totally “free”."
CentOS and Debian are totally free and very good server
platforms. CentOS/RHEL are even derived from Fedora, which means that
Fedora can't be that bad of a server platform.
--
Axel.Thimm at ATrpms.net
2:47 p.m.
Arthur Pemberton wrote:
> Of course they do, I'd be surprised to hear if anyone has
shipped a 100%
> bug free kernel, whether it be Linux, Solaris, AIX, Windows or whatever.
> The point though is that RHEL is less of a moving target, if you current
> setup works then an update to the kernel is less likely to break
> anything, than say Fedora which frequently ships new versions of the
> kernel.
One would hope that that these people complaining abut imperfect
kernels in RHEL are find these bugs on their test boxes _before_ they
deploy to production.
Absolutely, testing is good practice (regardless of your server OS
environment) before live deployment. Having said that, even the most
diligent of sysadmins can be caught out occasionally, but it stands to
reason that a new kernel update which contains a couple of bug fixes to
the same basic version is likely to present fewer surprises than going
from 2.6.23 -> 2.6.24 for example.
> > And the second you add a driver and/or XFS on to RHEL5 you
are
> > now tainted and *UNSUPPORTED*.
> Compared to Fedora where you are *UNSUPPORTED* at the offset?
is XFS even in the vanilla kernel?
I believe it is, I did a quick search on google and ironically it brough
up a post about xfs oopsing in 2.6.24 :-P
--
Ian Chapman.
2:47 p.m.
Roger Heflin wrote:
>>
>> Fedora ships buggy kernels. As stated previously this is more
>> upstream's fault,
>
> Beg your pardon, but just because someone writes broken code that
> _does not_ force a distribution to ship it.
>
The problem is that the enterprise OS's ship buggy kernels too,
Yes, but they very, very rarely introduce new bugs in an update that
breaks a previously working system.
I have
found at least one bug in most of the recent enterprise kernels (RHEL3,
RHEL4, RHEL5 and SLES9-never tested SLES10), and some of those bugs were
very very ugly had at least one of them had long since been fixed
upstream, and at least one of those bugs *NEVER* existed in the
kernel.org kernels at all.
And the timeline between reporting the bug and getting the fix was...?
And lets take RHEL5 initial release, it did not support file systems
larger then 8TB (ext3 only, no XFS), it did not support Areca and 3ware
PCIe controllers even though those drivers had been out for 6+ months at
the time they shipped RHEL5, and those are most definitely enterprise
boards. And the second you add a driver and/or XFS on to RHEL5 you are
now tainted and *UNSUPPORTED*.
Do you have some point here - like a system that always supports
everything and never fails? I usually don't have as much of a problem
having to fiddle with a new machine where you obviously aren't already
relying on it or getting known-supported hardware when I don't have time
to fiddle as I do with updates breaking previously working things. But,
have you tried the Centosplus kernels that put xfs and the drivers that
rhel removes back?
--
Les Mikesell
lesmiksell(a)gmail.com
2:25 p.m.
Roger Heflin wrote:
And lets take RHEL5 initial release, it did not support file systems
larger then 8TB (ext3 only, no XFS), it did not support Areca and 3ware
PCIe controllers even though those drivers had been out for 6+ months at
the time they shipped RHEL5, and those are most definitely enterprise
boards. And the second you add a driver and/or XFS on to RHEL5 you are
now tainted and *UNSUPPORTED*.
This is not quite the whole picture. If a third party kernel module is
loaded, that module is specifically unsupported assuming a ISV is not
involved. If a problem occurs that is not related to the module and you
can demonstrate that by reproducing the problem after unloading it,
support must be provided.
Rahul
2:24 p.m.
On Tue, Mar 25, 2008 at 2:17 PM, Ian Chapman
<packages(a)amiga-hardware.com> wrote:
Roger Heflin wrote:
>
> The problem is that the enterprise OS's ship buggy kernels too, I have
Of course they do, I'd be surprised to hear if anyone has shipped a 100%
bug free kernel, whether it be Linux, Solaris, AIX, Windows or whatever.
The point though is that RHEL is less of a moving target, if you current
setup works then an update to the kernel is less likely to break
anything, than say Fedora which frequently ships new versions of the
kernel.
One would hope that that these people complaining abut imperfect
kernels in RHEL are find these bugs on their test boxes _before_ they
deploy to production.
> And the second you add a driver and/or XFS on to RHEL5 you are
> now tainted and *UNSUPPORTED*.
Compared to Fedora where you are *UNSUPPORTED* at the offset?
is XFS even in the vanilla kernel?
--
Fedora 7 : sipping some of that moonshine
( www.pembo13.com )
2:23 p.m.
David G. Mackay wrote:
Now, if we're talking a system that's handling financial transactions
that has to be available 24/7, then it's another situation entirely.
I assumed that the word 'servers' in the subject meant that the services
mattered to someone...
I've been in shops like that, and am aware of the practices
necessary to
sustain that. I don't, however, need a disaster recovery plan with a
hot backup site for my video and music collection.
If reliability over some long interval isn't important to you, then this
thread probably isn't very interesting.
--
Les Mikesell
lesmikesell(a)gmail.com
2:17 p.m.
Roger Heflin wrote:
The problem is that the enterprise OS's ship buggy kernels too, I have
Of course they do, I'd be surprised to hear if anyone has shipped a 100%
bug free kernel, whether it be Linux, Solaris, AIX, Windows or whatever.
The point though is that RHEL is less of a moving target, if you current
setup works then an update to the kernel is less likely to break
anything, than say Fedora which frequently ships new versions of the
kernel.
And the second you add a driver and/or XFS on to RHEL5 you are
now tainted and *UNSUPPORTED*.
Compared to Fedora where you are *UNSUPPORTED* at the offset?
--
Ian Chapman.
1:48 p.m.
2008/3/25 max bianco <maximilianbianco(a)gmail.com>:
On Tue, Mar 25, 2008 at 1:34 PM, Les Mikesell <lesmikesell(a)gmail.com> wrote:
>
>
> Albert Graham wrote:
> >
> >>> I have installed hundreds of servers using Fedora and I have to say
> >>> I've had very few problems, kernel issues are not really the fault
of
> >>> the Fedora team, sometimes you hit quirks but these do get sorted out.
> >>
> >> How many years have you maintained these servers and how much damage
> >> does downtime cause?
> >>
> > None, as they are almost all clustered/load balanced/redundant.
>
>
> OK, but you might have mentioned that in your first post which could
> have been taken to mean hundreds of different offices were each relying
> on the one server you set up there. Fedora is OK if downtime doesn't
> matter.
>
>
>
> > Originally I was using RH 2.1 then 3, however, I found myself constantly
> > upgrading components because RH did not want to break "version"
> > compatibility for 5 years, which in my eyes is worse than binary
> > compatibly - Moores law and all that! so Fedora suits me down to the
> > ground.
> >
> > The only real issue is a stable kernel for your requirements, everything
> > else is less important, also I have a habit of running everything in
> > user-space so it's a lot easier to virtualize or switch out the
> > underlying OS if required.
>
>
> How do you virtualize "everything"? You have to have a real kernel and
> device drivers somewhere. If that isn't an up-to-date fedora then you
> are talking about something very different than it seems here.
>
>
> --
>
> Les Mikesell
> lesmikesell(a)gmail.com
>
>
am I the only person that voluntarily uses Fedora?
No you're not alone. A lot of us use it with no (major) issue what so
ever and so have no need to troll the mailing lists and complain.
My most recently "Fedora problem" was hardware related and due to
excessive dust build up in the machine, aka my fault.
Other problems I have had with Fedora are solely related to HAL and
UDEV, every now and then it fails to noticed a tv card of mine on a
cold boot. And now and then udev changes the assigned device name to
one of my two tv cards.
Thats it. I will note however that I am still on F7 due to lazyness
and lack of time. But I am not yet noticing any bit rot.
--
Fedora 7 : sipping some of that moonshine
( www.pembo13.com )
1:47 p.m.
On Tue, 2008-03-25 at 10:07 -0500, Les Mikesell wrote:
There was a point in mid-FC5 where an update kernel would not boot on
an
IBM Xseries 225 (fairly mainstream dual-xeon boxes, I thought) without
doing a motherboard bios update - after which the older kernels wouldn't
boot. And of course there were no warnings about this or much of a
reason to expect it to work after doing the bios flash either. And
there have been several updates that failed to boot on popular Dell and
IBM MPT scsi controllers - even one late into FC6 which was otherwise
pretty stable.
I'm not arguing that it doesn't happen. I've experienced a few. The
original post was directed to a fellow who wants to run a home server on
his desktop. And, I haven't seen anything yet that would cause more
than a few minutes downtime while switching to the backup partition
after noticing the problem.
Now, if we're talking a system that's handling financial transactions
that has to be available 24/7, then it's another situation entirely.
I've been in shops like that, and am aware of the practices necessary to
sustain that. I don't, however, need a disaster recovery plan with a
hot backup site for my video and music collection.
Dave
3:11 p.m.
On Mon, Mar 24, 2008 at 2:55 PM, Tim Alberts <talberts(a)msiscales.com> wrote:
Honestly I feel like a backstabber by using CentOS because I've
been
with Fedora since before Fedora (RedHat 5.1 was my first Linux
experience), and using CentOS is reaping the benefits with no contribution.
You realize that technically speaking Redhat|Centos are infact Fedora,
albeit mature versions of it.
--
Fedora 7 : sipping some of that moonshine
( www.pembo13.com )
1:31 p.m.
max bianco wrote:
am I the only person that voluntarily uses Fedora? I have only been
using it a short time but i understood it was a moving target when i
signed on. This thread reminds me of KL. It wasn't that long ago so i
am sure most of you remember to whom i am referring.
Fedora is what it is and some have more experience than others. Is there
some other lesson to take away from this thread?
Fedora tries to stay very close to current software development. That
means you get the new features as well as the new bugs. On a desktop or
as a second, non-critical machine that is a reasonable tradeoff,
especially if you use the applications where the new features matter.
This thread is mostly about servers where downtime is to be avoided at
(almost) any cost and you want things to keep working for years. It
isn't about fedora being inherently bad, it just isn't suitable for some
uses.
--
Les Mikesell
lesmikesell(a)gmail.com
1:23 p.m.
Albert Graham wrote:
I'm sorry that you don't understand my response,
Did you expect anyone to understand that your installations weren't typical?
but I did actually take
the time to visit the reference website you posted that "slated" the
Fedora a distribution, It was clear to me that, that guy does not have a
clue and is getting what he deserves, surely you understand that you
should not be taking his advice - that would be a case of the blind
leading the blind.
I didn't make the initial post with the reference. My comment was that
his seemed like a typical experience.
>>
>> The only real issue is a stable kernel for your requirements,
>> everything else is less important, also I have a habit of running
>> everything in user-space so it's a lot easier to virtualize or switch
>> out the underlying OS if required.
>
> How do you virtualize "everything"? You have to have a real kernel
> and device drivers somewhere. If that isn't an up-to-date fedora then
> you are talking about something very different than it seems here.
And I'd still like to understand this part. Do you only recommend
fedora for people who can virtualize it?
--
Les Mikesell
lesmikesell(a)gmail.com
1:03 p.m.
Hi Les,
I'm sorry that you don't understand my response, but I did actually take
the time to visit the reference website you posted that "slated" the
Fedora a distribution, It was clear to me that, that guy does not have a
clue and is getting what he deserves, surely you understand that you
should not be taking his advice - that would be a case of the blind
leading the blind.
Best regards,
Albert.
Les Mikesell wrote:
Albert Graham wrote:
>
>>> I have installed hundreds of servers using Fedora and I have to say
>>> I've had very few problems, kernel issues are not really the fault
>>> of the Fedora team, sometimes you hit quirks but these do get
>>> sorted out.
>>
>> How many years have you maintained these servers and how much damage
>> does downtime cause?
>>
> None, as they are almost all clustered/load balanced/redundant.
OK, but you might have mentioned that in your first post which could
have been taken to mean hundreds of different offices were each
relying on the one server you set up there. Fedora is OK if downtime
doesn't matter.
> Originally I was using RH 2.1 then 3, however, I found myself
> constantly upgrading components because RH did not want to break
> "version" compatibility for 5 years, which in my eyes is worse than
> binary compatibly - Moores law and all that! so Fedora suits me down
> to the ground.
>
> The only real issue is a stable kernel for your requirements,
> everything else is less important, also I have a habit of running
> everything in user-space so it's a lot easier to virtualize or switch
> out the underlying OS if required.
How do you virtualize "everything"? You have to have a real kernel
and device drivers somewhere. If that isn't an up-to-date fedora then
you are talking about something very different than it seems here.
12:34 p.m.
Albert Graham wrote:
>> I have installed hundreds of servers using Fedora and I have to say
>> I've had very few problems, kernel issues are not really the fault of
>> the Fedora team, sometimes you hit quirks but these do get sorted out.
>
> How many years have you maintained these servers and how much damage
> does downtime cause?
>
None, as they are almost all clustered/load balanced/redundant.
OK, but you might have mentioned that in your first post which could
have been taken to mean hundreds of different offices were each relying
on the one server you set up there. Fedora is OK if downtime doesn't
matter.
Originally I was using RH 2.1 then 3, however, I found myself
constantly
upgrading components because RH did not want to break "version"
compatibility for 5 years, which in my eyes is worse than binary
compatibly - Moores law and all that! so Fedora suits me down to the
ground.
The only real issue is a stable kernel for your requirements, everything
else is less important, also I have a habit of running everything in
user-space so it's a lot easier to virtualize or switch out the
underlying OS if required.
How do you virtualize "everything"? You have to have a real kernel and
device drivers somewhere. If that isn't an up-to-date fedora then you
are talking about something very different than it seems here.
--
Les Mikesell
lesmikesell(a)gmail.com
11:24 a.m.
Timothy Murphy wrote:
>> I haven't noticed any difference between CentOS and Fedora-8,
>> which I'm running on my other machines.
>> In fact, I don't know which I am on (when accessing remotely)
>> except by looking at the name.
> Based on past experience, you could expect the CentOS box to keep
> running with nothing but occasional 'yum update's for 6 more years.
True (well, true after s/6/2/), but irrelevant to the point I was making.
From http://wiki.centos.org/FAQ/CentOS5
"We intend to support CentOS 5 until Mar 31st, 2014"
After 2010 you only get security/maintenance updates, though.
> Even
> if F8 updates don't actually break your working system (which is likely
> to depend on your specific hardware) in less than a year you'll stop
> getting updates and as soon as a new security issue is discovered you'll
> be forced to re-install some wildly different version.
Personally, I keep both CentOS and Fedora up to date.
No Fedora updates have ever "broken" my system,
and I have been running Fedora since it started.
No firewire? No MPT scsi? No CIPE VPNs?
I've never been "forced to re-install some wildly different
version".
The FC1->2, FC3-4, FC6-7 transitions had huge differences. If you don't
use any of the features you might not notice the difference. Try
relying on backups on firewire drives (broken in FC3, most of FC5, or
connecting offices with CIPE vpn's (gone since FC1..). The
disk/partition names have changed, filesystem options have changed such
that creating an ext3 file with current options makes something an older
system can't read. Apache's mod_perl that was finally integrated
correctly back in RH7.3 has been a mixed bag in fedora through FC6.
Sound has been hit-or-miss with huge differences in each version. Fixing
any of these things can take days - which may not sound bad but if you
manage a few hundred machines, you can't keep up with a few days per
machine per year.
--
Les Mikesell
lesmikesell(a)gmail.com
11:21 a.m.
Les Mikesell wrote:
Albert Graham wrote:
>
> I have installed hundreds of servers using Fedora and I have to say
> I've had very few problems, kernel issues are not really the fault of
> the Fedora team, sometimes you hit quirks but these do get sorted out.
How many years have you maintained these servers and how much damage
does downtime cause?
None, as they are almost all clustered/load balanced/redundant.
Originally I was using RH 2.1 then 3, however, I found myself constantly
upgrading components because RH did not want to break "version"
compatibility for 5 years, which in my eyes is worse than binary
compatibly - Moores law and all that! so Fedora suits me down to the ground.
The only real issue is a stable kernel for your requirements, everything
else is less important, also I have a habit of running everything in
user-space so it's a lot easier to virtualize or switch out the
underlying OS if required.
Albert.
10:42 a.m.
Les Mikesell wrote:
Mauriat M wrote:
>
> Fedora ships buggy kernels. As stated previously this is more
> upstream's fault,
Beg your pardon, but just because someone writes broken code that _does
not_ force a distribution to ship it.
The problem is that the enterprise OS's ship buggy kernels too, I have found at
least one bug in most of the recent enterprise kernels (RHEL3, RHEL4, RHEL5 and
SLES9-never tested SLES10), and some of those bugs were very very ugly had at
least one of them had long since been fixed upstream, and at least one of those
bugs *NEVER* existed in the kernel.org kernels at all.
And lets take RHEL5 initial release, it did not support file systems larger then
8TB (ext3 only, no XFS), it did not support Areca and 3ware PCIe controllers
even though those drivers had been out for 6+ months at the time they shipped
RHEL5, and those are most definitely enterprise boards. And the second you add
a driver and/or XFS on to RHEL5 you are now tainted and *UNSUPPORTED*.
Roger
10:40 a.m.
Ed Greshko wrote:
>>
>> Fedora ships buggy kernels. As stated previously this is more
>> upstream's fault,
>
> Beg your pardon, but just because someone writes broken code that
> _does not_ force a distribution to ship it.
Of course it does not force a distro to ship it...but if the distro
expects that it will be handed "good code" and ships on that basis
then.....
Sorry, what is your point?
Fedora ships stuff with little or no testing. More realistically,
fedora users _do_ the testing of new kernel development. Personally, I
think every linux distro should have refused to ship 2.6 until a new
development branch was started. 2.2 and 2.4 became fairly stable at
around their x.x.20 releases because the experimental work was done in
an odd numbered development branch. I don't foresee 2.6 ever
stabilizing the way it is currently handled and it is entirely up to the
distros to sort out what is usable. This is good for the for-pay
enterprise distributions, of course, since it makes the others less
useful for anything where dependability matters.
--
Les Mikesell
lesmikesell(a)gmail.com
10:15 a.m.
Les Mikesell wrote:
>> Being on the leading edge mitigates seriously against
reliability.
>
> I find the boast/warning that Fedora is "bleeding edge" slightly absurd.
>
> In my (probably limited) experience, it is neither more nor less reliable
> than other distributions.
Then perhaps you should listen to those with more experience.
I'd listen to anybody who gives some sort of evidence for their views.
Bald assertions are not worth the paper they are not written on.
Eg in this case it would be more useful
if you actually specified some distribution which in your experience
is more reliable that Fedora, rather than simply asserting
(as you seem to) that you have more experience than me.
> I haven't noticed any difference between CentOS and
Fedora-8,
> which I'm running on my other machines.
> In fact, I don't know which I am on (when accessing remotely)
> except by looking at the name.
Based on past experience, you could expect the CentOS box to keep
running with nothing but occasional 'yum update's for 6 more years.
True (well, true after s/6/2/), but irrelevant to the point I was making.
Even
if F8 updates don't actually break your working system (which is likely
to depend on your specific hardware) in less than a year you'll stop
getting updates and as soon as a new security issue is discovered you'll
be forced to re-install some wildly different version.
Personally, I keep both CentOS and Fedora up to date.
No Fedora updates have ever "broken" my system,
and I have been running Fedora since it started.
I've never been "forced to re-install some wildly different version".
--
Timothy Murphy
e-mail (<80k only): tim /at/ birdsnest.maths.tcd.ie
tel: +353-86-2336090, +353-1-2842366
s-mail: School of Mathematics, Trinity College, Dublin 2, Ireland
10:07 a.m.
David G. Mackay wrote:
On Tue, 2008-03-25 at 20:51 +0900, John Summerfield wrote:
> There's losing and there's really losing. If Fedora won't boot, you have
> the services of neither until it's fixed.
>
> A little shore of Best Practice.
I don't think that anyone was talking about HA-Linux and redundant power
supplies here, John. It's fairly trivial to set up a separate partition
with a fallback OS. I routinely do this with Fedora on one of my
machines, and then rotate versions of Fedora between the two partitions
as the releases become available. If an update makes the newer version
unbootable (extremely rare that you can't just use grub to boot the
previous kernel), then there's always the old version to boot into.
There was a point in mid-FC5 where an update kernel would not boot on an
IBM Xseries 225 (fairly mainstream dual-xeon boxes, I thought) without
doing a motherboard bios update - after which the older kernels wouldn't
boot. And of course there were no warnings about this or much of a
reason to expect it to work after doing the bios flash either. And
there have been several updates that failed to boot on popular Dell and
IBM MPT scsi controllers - even one late into FC6 which was otherwise
pretty stable.
--
Les Mikesell
lesmikesell(a)gmail.com
10:06 a.m.
Les Mikesell wrote:
Mauriat M wrote:
>
> Fedora ships buggy kernels. As stated previously this is more
> upstream's fault,
Beg your pardon, but just because someone writes broken code that _does
not_ force a distribution to ship it.
Of course it does not force a distro to ship it...but if the distro expects
that it will be handed "good code" and ships on that basis then.....
Sorry, what is your point?
9:58 a.m.
On Tue, Mar 25, 2008 at 10:58 AM, Les Mikesell <lesmikesell(a)gmail.com> wrote:
Mauriat M wrote:
>
> Fedora ships buggy kernels. As stated previously this is more
> upstream's fault,
Beg your pardon, but just because someone writes broken code that _does
not_ force a distribution to ship it.
I totally agree, that is what I was hoping people would realize. To
clarify: the "buggy-ness" of the kernel is not to be blamed on Fedora
directly.
-Mauriat
9:53 a.m.
Mauriat M wrote:
> this guy should be using RHEL if he does not want anything to change.
So RHEL is the only way someone might want to have a desktop linux
that can in some capacity act as a server. Noted.
Or CentOS, if you don't mind your updates coming a few hours later.
--
Les Mikesell
lesmikesell(a)gmail.com
9:40 a.m.
On Tue, Mar 25, 2008 at 10:03 AM, Timothy Murphy <gayleard(a)eircom.net> wrote:
I find the boast/warning that Fedora is "bleeding edge" slightly absurd.
Some may find the notion that Fedora is "stable" slightly absurd.
In my (probably limited) experience, it is neither more nor less
reliable
than other distributions.
Your experience is your own, as this list so abundantly makes clear
each has their own experience.
The complaint in the posting that started this,
<http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/>;,
seemed to be about some unspecified problem with ssh, under some kernel.
I don't recall anyone else mentioning this.
I certainly didn't notice any problem.
Fedora ships buggy kernels. As stated previously this is more
upstream's fault, however that being said, the kernel is the single
most critical aspect of the operating system. When it is buggy the
entire system is rather, ummm, buggy.
As for the TCP bug it seemed to be with every post 2.6.21 kernel.
While it helps me little if others have noticed it, it was 100%
reproducible for me. Since the information to me indicated this was
"fixed upstream", I just waited, which is most likely what I would
have been told to do. I have not recently tested the later 23 kernels
but last I checked the changelogs from kernel.org didn't reveal any
useful information.
I have been meaning to try the 24 kernels, but I was waiting till the
storm of "2.6.24 borked my system" was over.
I'll do some testing over the next few days. Its not fun doing "remote"
testing.
I've actually gone over to CentOS-5.1 on my tiny server,
but that is because it is running on a Dell PowerEdge T105
(which for some reason Dell is or was selling for 150 euro)
and Dell lists CentOS among its 2 or 3 "acceptable OSes".
This means you get official Dell software and hardware related updates.
I haven't noticed any difference between CentOS and Fedora-8,
which I'm running on my other machines.
Perhaps you haven't looked hard enough, perhaps you don't need to.
In fact, I don't know which I am on (when accessing remotely)
except by looking at the name.
In any event I'm almost done migrating all "server" related functions
to a separate CentOS machine. That being said I find this whole thread
a little amusing. Just a random rant on a random website by a random
Fedora user. Nothing new here, move along.
-Mauriat
9:51 a.m.
Timothy Murphy wrote:
>
> Being on the leading edge mitigates seriously against reliability.
I find the boast/warning that Fedora is "bleeding edge" slightly absurd.
In my (probably limited) experience, it is neither more nor less reliable
than other distributions.
Then perhaps you should listen to those with more experience.
I haven't noticed any difference between CentOS and Fedora-8,
which I'm running on my other machines.
In fact, I don't know which I am on (when accessing remotely)
except by looking at the name.
Based on past experience, you could expect the CentOS box to keep
running with nothing but occasional 'yum update's for 6 more years. Even
if F8 updates don't actually break your working system (which is likely
to depend on your specific hardware) in less than a year you'll stop
getting updates and as soon as a new security issue is discovered you'll
be forced to re-install some wildly different version.
--
Les Mikesell
lesmikesell(a)gmail.com
9:49 a.m.
On Tue, Mar 25, 2008 at 9:40 AM, Mauriat M <mirandam(a)gmail.com> wrote:
On Tue, Mar 25, 2008 at 10:03 AM, Timothy Murphy
<gayleard(a)eircom.net> wrote:
>
> I find the boast/warning that Fedora is "bleeding edge" slightly
absurd.
Some may find the notion that Fedora is "stable" slightly absurd.
> In my (probably limited) experience, it is neither more nor less reliable
> than other distributions.
Your experience is your own, as this list so abundantly makes clear
each has their own experience.
This is a list where people with problems come for help. It is
entirely possible that this list only represents 1% of the Fedora
usage, I have no idea.
--
Fedora 7 : sipping some of that moonshine
( www.pembo13.com )
9:37 a.m.
On Mon, Mar 24, 2008 at 5:01 PM, Albert Graham <agraham(a)g-b.net> wrote:
You get out of it what you put in it, this guy clearly could not be
bothered to look into issues that he was having or why things had
changed - which is called progress.
If something "changes" and that change breaks something badly should
it be called progress?
I have installed hundreds of servers using Fedora and I have to say
I've
had very few problems, kernel issues are not really the fault of the
Fedora team, sometimes you hit quirks but these do get sorted out.
If you haven't already, go read the fedora-list or fedora-devel
archives on the 2.6.24 kernel.
I find the path / progress and choices Fedora makes are
"natural
progression" and indeed ahead of the pack.
I find that a opinion is just an opinion.
this guy should be using RHEL if he does not want anything to
change.
So RHEL is the only way someone might want to have a desktop linux
that can in some capacity act as a server. Noted.
Does not want to break is independent of does not want to change.
My 2c.
Which is probably the same value as the rant on that webpage.
-Mauriat
9:03 a.m.
John Summerfield wrote:
Perhaps you should read this
document,http://fedoraproject.org/wiki/Objectives and particularly note
this para. "To be on the leading edge of free and open source
technology, by adopting and helping develop new features and version
upgrades."
Being on the leading edge mitigates seriously against reliability.
I find the boast/warning that Fedora is "bleeding edge" slightly absurd.
In my (probably limited) experience, it is neither more nor less reliable
than other distributions.
The complaint in the posting that started this,
<http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/>;,
seemed to be about some unspecified problem with ssh, under some kernel.
I don't recall anyone else mentioning this.
I certainly didn't notice any problem.
I've actually gone over to CentOS-5.1 on my tiny server,
but that is because it is running on a Dell PowerEdge T105
(which for some reason Dell is or was selling for 150 euro)
and Dell lists CentOS among its 2 or 3 "acceptable OSes".
This means you get official Dell software and hardware related updates.
I haven't noticed any difference between CentOS and Fedora-8,
which I'm running on my other machines.
In fact, I don't know which I am on (when accessing remotely)
except by looking at the name.
--
Timothy Murphy
e-mail (<80k only): tim /at/ birdsnest.maths.tcd.ie
tel: +353-86-2336090, +353-1-2842366
s-mail: School of Mathematics, Trinity College, Dublin 2, Ireland
9:03 a.m.
On Tue, 2008-03-25 at 20:51 +0900, John Summerfield wrote:
There's losing and there's really losing. If Fedora won't
boot, you have
the services of neither until it's fixed.
A little shore of Best Practice.
I don't think that anyone was talking about HA-Linux and redundant power
supplies here, John. It's fairly trivial to set up a separate partition
with a fallback OS. I routinely do this with Fedora on one of my
machines, and then rotate versions of Fedora between the two partitions
as the releases become available. If an update makes the newer version
unbootable (extremely rare that you can't just use grub to boot the
previous kernel), then there's always the old version to boot into.
Having a VM running a server in this environment may not be the absolute
best practice, but it's certainly feasible.
Dave
8:55 a.m.
On Tue, Mar 25, 2008 at 06:44:05 -0500,
Bradley Pursley <pursley001(a)comcast.net> wrote:
Another problem I've noticed is that when new software is being
installed, there are no tests done to verify that it installed and is
working correctly, which has been a big problem on my system. I've had
to manually "fix" quite a number of the installs and sometimes when it
updates the software, it undoes my fixes and I have to re-apply the fix
again, which is highly aggravating. Updates should not automatically
mess with the configuration files, especially if they've been manually
changed, unless it prompts the system user and gives you the option of
whether you want the updated file or just want to adjust the current
one. For programming, this would be easy to implement without changing
packaging requirements (don't overwrite older non-executable files, just
rename them or move them to a "backup" location and advise user of the
change).
Updates shouldn't be messing with config files in the vast majority of
the packages. Normally the new config file gets name whatever.rpmnew and
you need to manually look at what's changed and see if you need to do anything.
In the rare cases where the config file needs to be replaced, the old one
should be named whatever.rpmsave and you again need to manually look at
what changed and decide what to do. You can use a find command to find all
of these in /etc that will run pretty quick.
8:07 a.m.
John Summerfield wrote:
Arthur Pemberton wrote:
> On Mon, Mar 24, 2008 at 2:55 PM, Tim Alberts <talberts(a)msiscales.com>
> wrote:
>> Honestly I feel like a backstabber by using CentOS because I've been
>> with Fedora since before Fedora (RedHat 5.1 was my first Linux
>> experience), and using CentOS is reaping the benefits with no
>> contribution.
>
> You realize that technically speaking Redhat|Centos are infact Fedora,
> albeit mature versions of it.
>
Not really. The testing that goes into RHEL is quite different.
And yet there are packages in RHEL5 that weren't rebuilt from their FC6
versions...
--
Les Mikesell
lesmikesell(a)gmail.com
8:05 a.m.
Alan Cox wrote:
> The _only_ reason I will admit for its use as a server is for
software
> development, where one is developing client or server against it,
> targetting a future release of RHEL. Then, I would use it as needed for
> that development, but not for other purposes.
>
> Then I might run Fedora and Rawhide, both, for development, and test
> against both.
>
> Just read this list and see how often Fedora blows up.
I think we should at this point try 'statistics for beginners'. You have
a list consisting of people who are heavily involved and people who join
because they have problems combined with a tendancy for those having
problems to be the ones who post (why email 'my server is working today'
to the list).
There are lies, damn lies, and statistics... I'd say the people on this
list are far, far better than average at fixing their own problems and
they still end up posting about things that they can't get to work.
Products such as RHEL are designed with a goal of avoiding
regressions,
so what works continues to work and for many uses that is far more
preferable. To understand Fedora reliability you need to work out what to
measure. The end result of that is that while RHEL might be far less
likely to regress something that works well than Fedora, it is also far
less likely to cure something than Fedora.
I've only seen the 'cure' case where the hardware was newer than the
RHEL/Centos release that was loaded on it and fedora had a
correspondingly new driver. And updates for popular hardware are
usually backed into RHEL/Centos kernels.
--
Les Mikesell
lesmikesell(a)gmail.com
7:30 a.m.
Actually, CentOS 5.1 isn't that bad as a desktop. So you could probably get
away
with using just one machine. The Linux desctop have come a long way, and you
no
longer need to be on the bleeding edge to have a resonable set of
functionality.
2008/3/24, Mauriat M <mirandam(a)gmail.com>:
On Mon, Mar 24, 2008 at 3:53 PM, Arthur Pemberton <pemboa(a)gmail.com>
wrote:
> On Mon, Mar 24, 2008 at 2:33 PM, Valent Turkovic
> <valent.turkovic(a)gmail.com> wrote:
> >
http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/
> >
> > What is your experience ?
>
> The general wisdom from fedora power users is:
> - Fedora for the desktop / workstation
> - CentOS for the home server or other unsupported server
> - RedHat for a serious server or otherwise when support is needed
>
> * by support I mean paid support
>
> Also, SELinux isn't nearly that hard. I wouldn't put a box on the open
> internet without SELinux.
>
Well I guess I was hoping for the best of both worlds by having my
desktop "act" as a server. Why have 2 machines when 1 will do?
Too much to ask, I know.
-Mauriat
--
fedora-list mailing list
fedora-list(a)redhat.com
To unsubscribe: https://www.redhat.com/mailman/listinfo/fedora-list
7:11 a.m.
Bradley Pursley wrote:
Valent Turkovic wrote:
> http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/
>
> What is your experience ?
>
> Cheers,
> Valent.
My main concern / problem is that it doesn't seem like the updates are
Perhaps you should read this
document,http://fedoraproject.org/wiki/Objectives and particularly note
this para. "To be on the leading edge of free and open source
technology, by adopting and helping develop new features and version
upgrades."
Being on the leading edge mitigates seriously against reliability.
very well tested thereby *CAUSING* crashes, lockups, etc. Until this
lack of concern is fixed, I would not recommend Fedora for servers.
Something that might help is by allowing updates to be undone
automatically if there is a problem and then to "lock out" that update
from being applied again until the next generation of that update comes
out.
Another problem I've noticed is that when new software is being
installed, there are no tests done to verify that it installed and is
working correctly, which has been a big problem on my system. I've had
to manually "fix" quite a number of the installs and sometimes when it
updates the software, it undoes my fixes and I have to re-apply the fix
again, which is highly aggravating. Updates should not automatically
mess with the configuration files, especially if they've been manually
changed, unless it prompts the system user and gives you the option of
whether you want the updated file or just want to adjust the current
one. For programming, this would be easy to implement without changing
packaging requirements (don't overwrite older non-executable files, just
rename them or move them to a "backup" location and advise user of the
change).
Just my ten cents worth.
Bradley
You need to more properly understand what it is you're using. Alan (Cox)
is a software developer, he has a different perspective than I have. My
primary interest has been keeping systems running, and that requires a
more conservative approach to systems maintenance.
--
Cheers
John
-- spambait
1aaaaaaa(a)coco.merseine.nu Z1aaaaaaa(a)coco.merseine.nu
-- Advice
http://webfoot.com/advice/email.top.php
http://www.catb.org/~esr/faqs/smart-questions.html
http://support.microsoft.com/kb/555375
You cannot reply off-list:-)
7:04 a.m.
Alan Cox wrote:
G'day Alan
I don't see the point of an off-list cc, I'm on the list and the round
trip from here to RH and back is a few seconds.
> The _only_ reason I will admit for its use as a server is for
software
> development, where one is developing client or server against it,
> targetting a future release of RHEL. Then, I would use it as needed for
> that development, but not for other purposes.
>
> Then I might run Fedora and Rawhide, both, for development, and test
> against both.
>
> Just read this list and see how often Fedora blows up.
I think we should at this point try 'statistics for beginners'. You have
a list consisting of people who are heavily involved and people who join
because they have problems combined with a tendancy for those having
problems to be the ones who post (why email 'my server is working today'
to the list).
All that's true, but it happens fairly often that something, such as a
kernel update, renders Fedora unusable, and it does not help that the
default behaviour when installing a new kernel is to make the new kernel
active for the next boot.
I maintain servers remotely, and while I might reboot one in the
expectation that should there be a problem I have a cross-town journey,
I really don't want the new kernel to be booted until my say-so.
Products such as RHEL are designed with a goal of avoiding regressions,
so what works continues to work and for many uses that is far more
preferable. To understand Fedora reliability you need to work out what to
measure. The end result of that is that while RHEL might be far less
likely to regress something that works well than Fedora, it is also far
less likely to cure something than Fedora.
Just a few minutes ago I was reading about a problem fixed in CentOS
that Red Hat (apparently) wasn't interested in fixing.
I know a lot of people who run Fedora reliably on servers, including some
I'd have expected to be running enterprise products.
Presumably they know enough to not be surprised when it breaks.
--
Cheers
John
-- spambait
1aaaaaaa(a)coco.merseine.nu Z1aaaaaaa(a)coco.merseine.nu
-- Advice
http://webfoot.com/advice/email.top.php
http://www.catb.org/~esr/faqs/smart-questions.html
http://support.microsoft.com/kb/555375
You cannot reply off-list:-)
6:51 a.m.
David G. Mackay wrote:
On Tue, 2008-03-25 at 16:52 +0900, John Summerfield wrote:
> David G. Mackay wrote:
>> Not really. If you have sufficient resources (disk, memory, processing
>> power) on your desktop, then you can run fedora, and a virtual copy of
>> Centos (which runs just fine under kvm on F7). If you don't have
>> hardware assisted virtualization, you can still run vmware, qemu, xen,
>> etc.
> So when Fedora won't boot you lose both. Brilliant.
You have heard of rescue cds, I trust, and backups. Actual disk
corruption is extremely rare, these days, but if that were a concern you
could always put your server image on a seperate partition. The chances
of actually losing something are pretty minimal. Down time could be a
bit higher, especially as you tend to reboot fedora systems to install
new kernels or hal, etc.
There's losing and there's really losing. If Fedora won't boot, you have
the services of neither until it's fixed.
A little shore of Best Practice.
--
Cheers
John
-- spambait
1aaaaaaa(a)coco.merseine.nu Z1aaaaaaa(a)coco.merseine.nu
-- Advice
http://webfoot.com/advice/email.top.php
http://www.catb.org/~esr/faqs/smart-questions.html
http://support.microsoft.com/kb/555375
You cannot reply off-list:-)
6:44 a.m.
Valent Turkovic wrote:
http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/
What is your experience ?
Cheers,
Valent.
My main concern / problem is that it doesn't seem like the updates are
very well tested thereby *CAUSING* crashes, lockups, etc. Until this
lack of concern is fixed, I would not recommend Fedora for servers.
Something that might help is by allowing updates to be undone
automatically if there is a problem and then to "lock out" that update
from being applied again until the next generation of that update comes
out.
Another problem I've noticed is that when new software is being
installed, there are no tests done to verify that it installed and is
working correctly, which has been a big problem on my system. I've had
to manually "fix" quite a number of the installs and sometimes when it
updates the software, it undoes my fixes and I have to re-apply the fix
again, which is highly aggravating. Updates should not automatically
mess with the configuration files, especially if they've been manually
changed, unless it prompts the system user and gives you the option of
whether you want the updated file or just want to adjust the current
one. For programming, this would be easy to implement without changing
packaging requirements (don't overwrite older non-executable files, just
rename them or move them to a "backup" location and advise user of the
change).
Just my ten cents worth.
Bradley
6:40 a.m.
On Tue, 2008-03-25 at 16:52 +0900, John Summerfield wrote:
David G. Mackay wrote:
> Not really. If you have sufficient resources (disk, memory, processing
> power) on your desktop, then you can run fedora, and a virtual copy of
> Centos (which runs just fine under kvm on F7). If you don't have
> hardware assisted virtualization, you can still run vmware, qemu, xen,
> etc.
So when Fedora won't boot you lose both. Brilliant.
You have heard of rescue cds, I trust, and backups. Actual disk
corruption is extremely rare, these days, but if that were a concern you
could always put your server image on a seperate partition. The chances
of actually losing something are pretty minimal. Down time could be a
bit higher, especially as you tend to reboot fedora systems to install
new kernels or hal, etc.
Dave
3:17 a.m.
The _only_ reason I will admit for its use as a server is for
software
development, where one is developing client or server against it,
targetting a future release of RHEL. Then, I would use it as needed for
that development, but not for other purposes.
Then I might run Fedora and Rawhide, both, for development, and test
against both.
Just read this list and see how often Fedora blows up.
I think we should at this point try 'statistics for beginners'. You have
a list consisting of people who are heavily involved and people who join
because they have problems combined with a tendancy for those having
problems to be the ones who post (why email 'my server is working today'
to the list).
Products such as RHEL are designed with a goal of avoiding regressions,
so what works continues to work and for many uses that is far more
preferable. To understand Fedora reliability you need to work out what to
measure. The end result of that is that while RHEL might be far less
likely to regress something that works well than Fedora, it is also far
less likely to cure something than Fedora.
I know a lot of people who run Fedora reliably on servers, including some
I'd have expected to be running enterprise products.
Alan
3:13 a.m.
On Mon, 2008-03-24 at 16:42 -0400, Mauriat M wrote:
Well I guess I was hoping for the best of both worlds by having my
desktop "act" as a server. Why have 2 machines when 1 will do?
There's lots of advantages in having 2 instead of 1. Quite apart from
being able to leave a simple server running all the time providing
services (public, or not), while you can fiddle all over the place with
your client machine and not lose anything important (because all that
important stuff is actually stored on the server). You've got two
machines for when you have to fault find. It's very handy to be able to
access the net on one machine, when a faulty one is refusing to do so.
--
(This computer runs FC7, my others run FC4, FC5 & FC6, in case that's
important to the thread.)
Don't send private replies to my address, the mailbox is ignored.
I read messages from the public lists.
2:53 a.m.
Arthur Pemberton wrote:
On Mon, Mar 24, 2008 at 2:55 PM, Tim Alberts
<talberts(a)msiscales.com> wrote:
> Honestly I feel like a backstabber by using CentOS because I've been
> with Fedora since before Fedora (RedHat 5.1 was my first Linux
> experience), and using CentOS is reaping the benefits with no contribution.
You realize that technically speaking Redhat|Centos are infact Fedora,
albeit mature versions of it.
Not really. The testing that goes into RHEL is quite different.
--
Cheers
John
-- spambait
1aaaaaaa(a)coco.merseine.nu Z1aaaaaaa(a)coco.merseine.nu
-- Advice
http://webfoot.com/advice/email.top.php
http://www.catb.org/~esr/faqs/smart-questions.html
http://support.microsoft.com/kb/555375
You cannot reply off-list:-)
2:52 a.m.
David G. Mackay wrote:
On Mon, 2008-03-24 at 16:42 -0400, Mauriat M wrote:
> Well I guess I was hoping for the best of both worlds by having my
> desktop "act" as a server. Why have 2 machines when 1 will do?
>
> Too much to ask, I know.
Not really. If you have sufficient resources (disk, memory, processing
power) on your desktop, then you can run fedora, and a virtual copy of
Centos (which runs just fine under kvm on F7). If you don't have
hardware assisted virtualization, you can still run vmware, qemu, xen,
etc.
So when Fedora won't boot you lose both. Brilliant.
--
Cheers
John
-- spambait
1aaaaaaa(a)coco.merseine.nu Z1aaaaaaa(a)coco.merseine.nu
-- Advice
http://webfoot.com/advice/email.top.php
http://www.catb.org/~esr/faqs/smart-questions.html
http://support.microsoft.com/kb/555375
You cannot reply off-list:-)
2:50 a.m.
Mauriat M wrote:
Well I guess I was hoping for the best of both worlds by having my
desktop "act" as a server. Why have 2 machines when 1 will do?
Too much to ask, I know.
CentOS
--
Cheers
John
-- spambait
1aaaaaaa(a)coco.merseine.nu Z1aaaaaaa(a)coco.merseine.nu
-- Advice
http://webfoot.com/advice/email.top.php
http://www.catb.org/~esr/faqs/smart-questions.html
http://support.microsoft.com/kb/555375
You cannot reply off-list:-)
2:49 a.m.
Valent Turkovic wrote:
Valent Turkovic wrote:
> http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/
>
> What is your experience ?
>
> Cheers,
> Valent.
>
My personal experience is that OpenSuse is a terrible server and I also
use CentOS now for my home server.
Valent.
Essentially the same thing, both OpenSuse and Fedora are rolling betas.
--
Cheers
John
-- spambait
1aaaaaaa(a)coco.merseine.nu Z1aaaaaaa(a)coco.merseine.nu
-- Advice
http://webfoot.com/advice/email.top.php
http://www.catb.org/~esr/faqs/smart-questions.html
http://support.microsoft.com/kb/555375
You cannot reply off-list:-)
2:48 a.m.
Valent Turkovic wrote:
http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/
What is your experience ?
I have not used it as a server, and do not plan to do so. For free
servers I use CentOS, a RHEL clone, and Debian.
The _only_ reason I will admit for its use as a server is for software
development, where one is developing client or server against it,
targetting a future release of RHEL. Then, I would use it as needed for
that development, but not for other purposes.
Then I might run Fedora and Rawhide, both, for development, and test
against both.
Just read this list and see how often Fedora blows up.
--
Cheers
John
-- spambait
1aaaaaaa(a)coco.merseine.nu Z1aaaaaaa(a)coco.merseine.nu
-- Advice
http://webfoot.com/advice/email.top.php
http://www.catb.org/~esr/faqs/smart-questions.html
http://support.microsoft.com/kb/555375
You cannot reply off-list:-)
10:24 p.m.
On Mon, 2008-03-24 at 16:42 -0400, Mauriat M wrote:
Well I guess I was hoping for the best of both worlds by having my
desktop "act" as a server. Why have 2 machines when 1 will do?
Too much to ask, I know.
Not really. If you have sufficient resources (disk, memory, processing
power) on your desktop, then you can run fedora, and a virtual copy of
Centos (which runs just fine under kvm on F7). If you don't have
hardware assisted virtualization, you can still run vmware, qemu, xen,
etc.
Dave
4:45 p.m.
On Mon, 2008-03-24 at 13:37 -0700, Tim Alberts wrote:
Fedora creates regular scheduled releases of a conglomeration of
selected projects on a regular schedule.
After the Fedora release has been run through the test/release cycle by
the Fedora users (us) it becomes the next version of RedHat (roughly).
Then CentOS takes the source packages of RedHat, puts their name on it
and re-distributes it (legally..).
So all the assembly work is done by Fedora, all the testing is done by
the Fedora users and all the rewards are taken by people who (only) use
CentOS while at least people who use RedHat pay for support.
Of course, some of us use Fedora as a main user machine to benefit from
all the whizzy new features and support for the latest gadgets, but then
use CentOS on the server that supports the user machines. The server
doesn't need to be too fancy, it should be running on stable, mature
hardware and bleeding-edge versions of things like Apache, dhcpd, NFS,
Samba or LDAP aren't required.
Dave
4:41 p.m.
On Mon, Mar 24, 2008 at 10:22 PM, Arthur Pemberton <pemboa(a)gmail.com> wrote:
On Mon, Mar 24, 2008 at 3:37 PM, Tim Alberts
<talberts(a)msiscales.com> wrote:
> Arthur Pemberton wrote:
> > On Mon, Mar 24, 2008 at 2:55 PM, Tim Alberts <talberts(a)msiscales.com>
wrote:
> >
> >> Honestly I feel like a backstabber by using CentOS because I've
been
> >> with Fedora since before Fedora (RedHat 5.1 was my first Linux
> >> experience), and using CentOS is reaping the benefits with no
contribution.
> >>
> >
> > You realize that technically speaking Redhat|Centos are infact Fedora,
> > albeit mature versions of it.
> >
> >
> Of course. What's your point?
That you need not feel like a backstabber
If you feel morally obligated then contribute in some way to Fedora
because that is contributing to RHEL and that is contributing to
CentOS so you don't feel like a backstabber :)
Valent.
--
http://kernelreloaded.blog385.com/
linux, blog, anime, spirituality, windsurf, wireless
registered as user #367004 with the Linux Counter, http://counter.li.org.
ICQ: 2125241, Skype: valent.turkovic
4:39 p.m.
On Mon, Mar 24, 2008 at 10:01 PM, Albert Graham <agraham(a)g-b.net> wrote:
Valent,
You get out of it what you put in it, this guy clearly could not be
bothered to look into issues that he was having or why things had
changed - which is called progress.
I have installed hundreds of servers using Fedora and I have to say I've
Why do you use Fedora and not CentOS for servers?
What is the benefit for you using Fedora for servers? How often do you
upgrade to latest version of fedora those hundereds of servers you
installed?
Valent.
--
http://kernelreloaded.blog385.com/
linux, blog, anime, spirituality, windsurf, wireless
registered as user #367004 with the Linux Counter, http://counter.li.org.
ICQ: 2125241, Skype: valent.turkovic
4:36 p.m.
Tim Alberts wrote:
>> Unfortunately, everything that is beneficial about Fedora
comes with
>> the price of 'not quite as well tested' status as RedHat or CentOS.
>>
>> Honestly I feel like a backstabber by using CentOS because I've been
>> with Fedora since before Fedora (RedHat 5.1 was my first Linux
>> experience), and using CentOS is reaping the benefits with no
>> contribution.
>
> If you go back that far, you should realize that fedora is
> approximately like the old RH X.0 releases (X from 4 to 7) and Centos
> is like the old RH X.2 or X.3 releases (free download of the tested
> and more stable releases). The numbering scheme is just different
> now. The for-pay-only RHEL is the part that diverged from the old
> scheme.
Yeah there are different groups filling the roles all along the
life-cycle of each release version.
You could even say that CentOS provides what FedoraLegacy intended to
provide.
Not really the point of what I'm saying though...
The point I was trying to make was that in the old days when RH made its
name by getting a community of users involved, the users who tested
the X.0 versions and reported the problems that made the fixes possible
were eventually given access to the X.2/X.3 versions containing those
fixes. Fedora/RH no longer works that way. The community putting up
with the X.0 problems never gets a stable update unless they switch to
Centos. The next fedora release will be like the old next X.0's were.
Again my question, how can Fedora produce a better tested product?
Fedora has the option of rebranding the Centos packages for a long-term
supported version or building their own similar version from the RHEL
sources. I think it would have been trivial to slipstream the FC1
updates to a copy of the Centos 3.x update repro, the FC3 -> Centos4,
and FC6 ->Centos5. In fact I think there are people who have done that
with yum even though it wasn't planned to work.
The
way it is, it's dis-respectful to the Fedora project for people to post
things like:
http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/
because it is the work of the Fedora Project and the users who end up
testing the software and suffering through software bugs and poorly
packaged projects.
I don't think so. Post something on the fedora developers site about
focusing on stability and you'll find no one is interested. In fact
they will point out their objectives:
http://fedoraproject.org/wiki/Objectives are to stay on the leading edge
and close to upstream development instead. It is not disrespectful to
say that they are meeting their objectives.
CentOS and RedHat would be no where near as stable
without Fedora.
No one says they would - but that doesn't make fedora a reasonable
server platform.
--
Les Mikesell
lesmikesell(a)gmail.com
4:22 p.m.
On Mon, Mar 24, 2008 at 3:37 PM, Tim Alberts <talberts(a)msiscales.com> wrote:
Arthur Pemberton wrote:
> On Mon, Mar 24, 2008 at 2:55 PM, Tim Alberts <talberts(a)msiscales.com> wrote:
>
>> Honestly I feel like a backstabber by using CentOS because I've been
>> with Fedora since before Fedora (RedHat 5.1 was my first Linux
>> experience), and using CentOS is reaping the benefits with no contribution.
>>
>
> You realize that technically speaking Redhat|Centos are infact Fedora,
> albeit mature versions of it.
>
>
Of course. What's your point?
That you need not feel like a backstabber
--
Fedora 7 : sipping some of that moonshine
( www.pembo13.com )
4:09 p.m.
Valent Turkovic wrote:
http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/
What is your experience ?
Cheers,
Valent.
I have been running a server using Fedora and before that the original
releases of RH. It is now running 8. I made a software RAID of 360GB and
use SAMBA to access it. I have two dozen systems that access it all the
time. Each revision of the OS is loaded with the RAID disconnected. I
update the distribution and connect the RAID. I configure SAMBA, usually
by copying the .conf file. I made a script that copies the user info. To
date I have never had a problem. I just added four 500GB drives in a
software RAID 5, again not a hitch.
I have a RSYNC box that keeps that server backed up. It has been running
Fedora core 2 since it was released. Again no problems.
I do not run anything else on either box and they have no other
peripherals defined, just the disk, LAN and console.
So, I don't have a problem using Fedora as a server.
4:05 p.m.
Mauriat M wrote:
On Mon, Mar 24, 2008 at 3:53 PM, Arthur Pemberton
<pemboa(a)gmail.com> wrote:
> On Mon, Mar 24, 2008 at 2:33 PM, Valent Turkovic
> <valent.turkovic(a)gmail.com> wrote:
> > http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/
> >
> > What is your experience ?
>
> The general wisdom from fedora power users is:
> - Fedora for the desktop / workstation
> - CentOS for the home server or other unsupported server
> - RedHat for a serious server or otherwise when support is needed
>
> * by support I mean paid support
>
> Also, SELinux isn't nearly that hard. I wouldn't put a box on the open
> internet without SELinux.
>
Well I guess I was hoping for the best of both worlds by having my
desktop "act" as a server. Why have 2 machines when 1 will do?
Too much to ask, I know.
You can do that, but you have to choose between stability and new
features. Server apps have mostly been feature-complete for years so
you avoid problems with the older app versions included in Centos. If
you are happy with those versions of desktop apps as well (about circa
FC6 in fedora terms) then you are all set. If you want newer desktop
apps, the way development currently works you have to take a new kernel
and associated unstable changes to get them.
--
Les Mikesell
lesmikesell(a)gmail.com
4:05 p.m.
Les Mikesell wrote:
Tim Alberts wrote:
>
> Unfortunately, everything that is beneficial about Fedora comes with
> the price of 'not quite as well tested' status as RedHat or CentOS.
>
> Honestly I feel like a backstabber by using CentOS because I've been
> with Fedora since before Fedora (RedHat 5.1 was my first Linux
> experience), and using CentOS is reaping the benefits with no
> contribution.
If you go back that far, you should realize that fedora is
approximately like the old RH X.0 releases (X from 4 to 7) and Centos
is like the old RH X.2 or X.3 releases (free download of the tested
and more stable releases). The numbering scheme is just different
now. The for-pay-only RHEL is the part that diverged from the old
scheme.
Yeah there are different groups filling the roles all along the
life-cycle of each release version.
You could even say that CentOS provides what FedoraLegacy intended to
provide.
Not really the point of what I'm saying though...
> Is there a way for Fedora to deal with more of the bugs before
> releasing and still remain a free distribution? I don't know how.
> If they do manage to do this, what would be the point of paying for
> RedHat?
You aren't supposed to be paying for the software in RHEL, you pay for
the support service.
<splithair/> Of course.
Again my question, how can Fedora produce a better tested product? The
way it is, it's dis-respectful to the Fedora project for people to post
things like:
http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/
because it is the work of the Fedora Project and the users who end up
testing the software and suffering through software bugs and poorly
packaged projects. CentOS and RedHat would be no where near as stable
without Fedora.
4:01 p.m.
Valent,
You get out of it what you put in it, this guy clearly could not be
bothered to look into issues that he was having or why things had
changed - which is called progress.
I have installed hundreds of servers using Fedora and I have to say I've
had very few problems, kernel issues are not really the fault of the
Fedora team, sometimes you hit quirks but these do get sorted out.
I find the path / progress and choices Fedora makes are "natural
progression" and indeed ahead of the pack.
this guy should be using RHEL if he does not want anything to change.
My 2c.
Albert.
Valent Turkovic wrote:
http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/
What is your experience ?
Cheers,
Valent.
3:57 p.m.
Valent Turkovic wrote:
http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/
What is your experience ?
Personally I run Fedora on our desktops which is perfect for that
environment, as users tend to demand or need the latest and greatest and
as our desktops tend to be updated once a year, the release cycle is
just about right. I wouldn't consider it for our servers though, the
release cycle is far too fast and changes happy too quickly. For that,
we use RHEL which tends to stay fairly static with the software supplied
with it, with bugfixes and security fixes tending to be back-ported. If
it wasn't for the fact we get the licenses for something like $35 then
we'd probably use CentOS.
--
Ian Chapman.
3:42 p.m.
On Mon, Mar 24, 2008 at 3:53 PM, Arthur Pemberton <pemboa(a)gmail.com> wrote:
On Mon, Mar 24, 2008 at 2:33 PM, Valent Turkovic
<valent.turkovic(a)gmail.com> wrote:
> http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/
>
> What is your experience ?
The general wisdom from fedora power users is:
- Fedora for the desktop / workstation
- CentOS for the home server or other unsupported server
- RedHat for a serious server or otherwise when support is needed
* by support I mean paid support
Also, SELinux isn't nearly that hard. I wouldn't put a box on the open
internet without SELinux.
Well I guess I was hoping for the best of both worlds by having my
desktop "act" as a server. Why have 2 machines when 1 will do?
Too much to ask, I know.
-Mauriat
3:38 p.m.
Tim Alberts wrote:
Unfortunately, everything that is beneficial about Fedora comes with the
price of 'not quite as well tested' status as RedHat or CentOS.
Honestly I feel like a backstabber by using CentOS because I've been
with Fedora since before Fedora (RedHat 5.1 was my first Linux
experience), and using CentOS is reaping the benefits with no contribution.
If you go back that far, you should realize that fedora is approximately
like the old RH X.0 releases (X from 4 to 7) and Centos is like the old
RH X.2 or X.3 releases (free download of the tested and more stable
releases). The numbering scheme is just different now. The
for-pay-only RHEL is the part that diverged from the old scheme.
Is there a way for Fedora to deal with more of the bugs before
releasing
and still remain a free distribution? I don't know how. If they do
manage to do this, what would be the point of paying for RedHat?
You aren't supposed to be paying for the software in RHEL, you pay for
the support service.
--
Les Mikesell
lesmikesell(a)gmail.com
3:37 p.m.
Arthur Pemberton wrote:
On Mon, Mar 24, 2008 at 2:55 PM, Tim Alberts
<talberts(a)msiscales.com> wrote:
> Honestly I feel like a backstabber by using CentOS because I've been
> with Fedora since before Fedora (RedHat 5.1 was my first Linux
> experience), and using CentOS is reaping the benefits with no contribution.
>
You realize that technically speaking Redhat|Centos are infact Fedora,
albeit mature versions of it.
Of course. What's your point?
Fedora creates regular scheduled releases of a conglomeration of
selected projects on a regular schedule.
After the Fedora release has been run through the test/release cycle by
the Fedora users (us) it becomes the next version of RedHat (roughly).
Then CentOS takes the source packages of RedHat, puts their name on it
and re-distributes it (legally..).
So all the assembly work is done by Fedora, all the testing is done by
the Fedora users and all the rewards are taken by people who (only) use
CentOS while at least people who use RedHat pay for support.
2:55 p.m.
Valent Turkovic wrote:
http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/
What is your experience ?
Cheers,
Valent.
I think I could have written that article myself. I couldn't agree
more. I've done the exact same thing, my servers are now CentOS, my
desktop is still Fedora.
I don't want to disrespect the Fedora community for all the work they do
because it is on the edge of the technology and it is very helpful to
all the projects the comprise Fedora to have so many people using the
software and providing feedback. Plus, how is open source software ever
going to become quality without lots of people testing/using/improving.
Unfortunately, everything that is beneficial about Fedora comes with the
price of 'not quite as well tested' status as RedHat or CentOS.
Honestly I feel like a backstabber by using CentOS because I've been
with Fedora since before Fedora (RedHat 5.1 was my first Linux
experience), and using CentOS is reaping the benefits with no contribution.
Is there a way for Fedora to deal with more of the bugs before releasing
and still remain a free distribution? I don't know how. If they do
manage to do this, what would be the point of paying for RedHat?
2:51 p.m.
Valent Turkovic wrote:
http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/
What is your experience ?
Cheers,
Valent.
My personal experience is that OpenSuse is a terrible server and I also
use CentOS now for my home server.
Valent.
2:50 p.m.
Valent Turkovic wrote:
http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/
What is your experience ?
That's pretty close to my experience and I'd throw in several kernel
updates that would not boot on my hardware and firewire drives that
weren't recognized for a year or so. But people usually want servers to
run for longer than fedora's expected life cycle anyway, so why even
attempt it?
--
Les Mikesell
lesmikesell(a)gmail.com
2:41 p.m.
Valent Turkovic wrote:
http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/
What is your experience ?
Cheers,
Valent.
Fedora is a operating system with a very short support time, that is,
updates will only be available for each release for around 1 year (when
Fedora 9 comes out Fedora 7 will stop receiving updates). So if you plan
on running your system longer then 1 year or so you should choose
something that has longer support cycle (RHEL, CentOS, Ubuntu LTS?).
My 2 cents ....
Finnur
1:12 p.m.
On Tue, Mar 25, 2008 at 1:34 PM, Les Mikesell <lesmikesell(a)gmail.com> wrote:
Albert Graham wrote:
>
>>> I have installed hundreds of servers using Fedora and I have to say
>>> I've had very few problems, kernel issues are not really the fault of
>>> the Fedora team, sometimes you hit quirks but these do get sorted out.
>>
>> How many years have you maintained these servers and how much damage
>> does downtime cause?
>>
> None, as they are almost all clustered/load balanced/redundant.
OK, but you might have mentioned that in your first post which could
have been taken to mean hundreds of different offices were each relying
on the one server you set up there. Fedora is OK if downtime doesn't
matter.
> Originally I was using RH 2.1 then 3, however, I found myself constantly
> upgrading components because RH did not want to break "version"
> compatibility for 5 years, which in my eyes is worse than binary
> compatibly - Moores law and all that! so Fedora suits me down to the
> ground.
>
> The only real issue is a stable kernel for your requirements, everything
> else is less important, also I have a habit of running everything in
> user-space so it's a lot easier to virtualize or switch out the
> underlying OS if required.
How do you virtualize "everything"? You have to have a real kernel and
device drivers somewhere. If that isn't an up-to-date fedora then you
are talking about something very different than it seems here.
--
Les Mikesell
lesmikesell(a)gmail.com
am I the only person that voluntarily uses Fedora? I have only been using it
a short time but i understood it was a moving target when i signed on.
This thread reminds me of KL. It wasn't that long ago so i am sure most of
you remember to whom i am referring.
Fedora is what it is and some have more experience than others. Is there
some other lesson to take away from this thread?
Max
8:06 a.m.
On Wed, 2008-03-26 at 00:58 -0600, Arthur Pemberton wrote:
> I was thinking of updating my firewall/router box to Centos
4.6. Do you
> have any thoughts/warnings on this?
>
>
>
> Dave
Why not Centos 5?
It's an AMD K6-2 processor. The effort to support that class of
processors on Centos 5.x seems to have stalled.
Dave
1:51 p.m.
On Tue, Mar 25, 2008 at 2:31 PM, Les Mikesell <lesmikesell(a)gmail.com> wrote:
max bianco wrote:
>
>
> am I the only person that voluntarily uses Fedora? I have only been
> using it a short time but i understood it was a moving target when i
> signed on. This thread reminds me of KL. It wasn't that long ago so i
> am sure most of you remember to whom i am referring.
> Fedora is what it is and some have more experience than others. Is there
> some other lesson to take away from this thread?
Fedora tries to stay very close to current software development. That
means you get the new features as well as the new bugs. On a desktop or
as a second, non-critical machine that is a reasonable tradeoff,
especially if you use the applications where the new features matter.
This thread is mostly about servers where downtime is to be avoided at
(almost) any cost and you want things to keep working for years. It
isn't about fedora being inherently bad, it just isn't suitable for some
uses.
I know it isn't suitable for some uses. I tried to learn what I could
about Fedora before I actually installed it on my machine. It seems
however that some people have unrealistic expectations. The bleeding
edge and stable are not words i would expect to fit into the same
sentence under any circumstances. Having said that it has been very stable
for me, and by stable i mean that it has never failed to boot, have i had
issues? Sure but the basic functionality has always been there for me. Maybe
some features break from time to time but i have always been able to access
the web to solve my problems, would i use it for a server? maybe but I'd not
have much room to complain if I tried to ride the bleeding edge and fell off
the cliff because of it.
7:25 a.m.
On Tuesday 25 March 2008 11:44:05 Bradley Pursley wrote:
Valent Turkovic wrote:
> http://www.mjmwired.net/linux/2008/02/11/fedora-makes-a-terrible-server/
>
> What is your experience ?
>
> Cheers,
> Valent.
My main concern / problem is that it doesn't seem like the updates are
very well tested thereby *CAUSING* crashes, lockups, etc. Until this
lack of concern is fixed, I would not recommend Fedora for servers.
Something that might help is by allowing updates to be undone
automatically if there is a problem and then to "lock out" that update
from being applied again until the next generation of that update comes
out.
Another problem I've noticed is that when new software is being
installed, there are no tests done to verify that it installed and is
working correctly, which has been a big problem on my system. I've had
to manually "fix" quite a number of the installs and sometimes when it
updates the software, it undoes my fixes and I have to re-apply the fix
again, which is highly aggravating. Updates should not automatically
mess with the configuration files, especially if they've been manually
changed, unless it prompts the system user and gives you the option of
whether you want the updated file or just want to adjust the current
one. For programming, this would be easy to implement without changing
packaging requirements (don't overwrite older non-executable files, just
rename them or move them to a "backup" location and advise user of the
change).
Just my ten cents worth.
Bradley
It seems that it depends on the complexity of either the server or the
services being run.
For example, I've run a redhat-and-then-fedora server for about 7 years. It
only uses apache, mysql, cups, ypserv, dovecot and serves my family's
enormous music and digital photograph collection. Fingrz-Cr0$$d as I type, it
has only crashed once and that was as a direct result of my fiddling with
lvm.
At work, I have a server that had run fedora 4 since this was the latest
version. It hasn't been patched since f4 stopped being supported and it never
crashed - in fact it was only re-booted twice in three years - when it needed
moving or extended electrical supply work was needed. I've upgraded it to f8
in the last week and now it only stays up for around 15 mins before dying
with no diagnostic. I'm hoping that yumming it up-to-date (presumably in 15
minute chunks!) will get rid of the problem.
Am I about to abandon Fedora server? Probably not, I'm a full-time teacher and
only a part-time system manager now. I don't have the time or the inclination
to cope with two distros, even if Centos is based on fedora. I just make sure
that I have a good backup and nothing critical on the horizon when I run yum!
My ha'pennyworth.
Nick
5856
days inactive
5877
days old
100 comments
30 participants
participants (30)
-
agraham -
Alan Cox -
Arthur Pemberton -
Axel Thimm -
Bradley -
Bruno Wolff III -
Da Rock -
David G. Mackay -
David Hough -
Ed Greshko -
Finnur Örn Guðmundsson -
Ian Chapman -
John Summerfield -
Les Mikesell -
Mauriat M -
max -
mjwestkamper -
Nicholas Robinson -
Patrick O'Callaghan -
Rahul Sundaram -
Roger Heflin -
Simon Andrews -
Strong -
Tim -
Tim Alberts -
Timothy Murphy -
Todd Denniston -
Tomasz Torcz -
Uno Engborg -
Valent Turkovic