On Mon, 2011-12-05 at 18:08 +0530, Soham Chakraborty wrote:
[Snip previous problem identification]
> This could be related to the
network card initializing itself later
> at the boot process while in the earlier steps, it has marked itself
> as active. I have seen incidents like this.
> Is it possible for you to
mount all the NFS mounted FS at the last
> stage of the boot process. Possibly creating a rc.local file, making
> it executable and do the NFS mounts from it. This can be one step of
> debugging.
> Another can be to simply add
_netdev option in the NFS exports and
> check whether it improves things.
> This is all what I can think
of now.
Well thanks for this. I think now that the NFS mounting may actually be
a red-herring - but you set me on to some more detailed troubleshooting.
I did 4 very unscientific tests. I started from a cold reboot and
measured the time from pressing the switch to the appearance of the GDM
login screen with a stopwatch on my phone (don't phone the Nobel prize
people just yet...). Tests were:
Test 1) As is - i.e. NFS mounts in fstab as previously described
Test 2) Not mounting the 3 NFS mounts at all (i.e. commenting out from fstab)
Test 3) Mounting NFS from fstab but with _netdev option
Test 4) Attempting to mount NFS drives from rc.local (note - I never got this to work)
Results:
Test 1: 0:59.2
Test 2: 0:57.8
Test 3: 1:54.2
Test 4: 0:58.9
(Note: All times include the 5s grub pause which I haven't removed yet)
However, something didn't seem right - after all the "as is" test seemed
just fine (under a minute including the 5s grub pause is quite
acceptable) but the long test 3 exhibited the behaviour I had noticed
before - the white tear-drop disappears for a while leaving an alarming
black screen - eventually returning as the blue "F" for Fedora logo and
then proceeding to the GDM screen.
So I repeated some tests:
Test 3 - repeat 1: 1:55.0
Test 1 - repeat 1: 1:58.7
Test 1 - repeat 2: 1:59.5
Test 1 - repeat 3: 0:56.4
Test 1 - repeat 4: 1:56.0
So the long boot times (c. 2 mins) seem to be random, given no changes
in configuration.
I looked at dmesg. I am no expert in what dmesg tells me, but here is
the last few lines of one of the long boots:
===============8<====================================
[root@localhost ~]# dmesg
[ snip ]
[ 32.248789] e1000e 0000:00:19.0: em1: 10/100 speed: disabling TSO
[ 32.248955] ADDRCONF(NETDEV_CHANGE): em1: link becomes ready
[ 42.354015] em1: no IPv6 routers present
[ 90.858903] SELinux: initialized (dev 0:21, type nfs4), uses genfs_contexts
[ 90.863622] SELinux: initialized (dev 0:22, type nfs4), uses genfs_contexts
[ 90.867429] SELinux: initialized (dev 0:23, type nfs4), uses genfs_contexts
[ 90.868787] SELinux: initialized (dev 0:24, type nfs4), uses genfs_contexts
[ 90.872920] SELinux: initialized (dev 0:22, type nfs4), uses genfs_contexts
[ 90.877188] SELinux: initialized (dev 0:25, type nfs4), uses genfs_contexts
[ 90.878883] mount[998]: mount.nfs: /mnt/NFSmark is busy or already mounted
[ 90.887451] SELinux: initialized (dev 0:22, type nfs4), uses genfs_contexts
[ 90.927896] SELinux: initialized (dev 0:21, type nfs4), uses genfs_contexts
[ 90.932400] SELinux: initialized (dev 0:22, type nfs4), uses genfs_contexts
[ 92.714872] [drm:drm_debugfs_create_files] *ERROR* Cannot create
/sys/kernel/debug/dri/channel\xfffffff0i\xffffff80]\xffffffc0\x03/3
[ 95.940096] [drm:drm_debugfs_create_files] *ERROR* Cannot create
/sys/kernel/debug/dri/channel\xffffffe9i\xffffff80]\xffffffc0\x03/3
[root@localhost ~]#
===============8<====================================
On the successful boots dmesg ends with the "no IPv6 routers present"
line at around 44s.
What is this telling me and how do I fix it?
Thanks again...
Mark