Greetings
I thought I would send out my thoughts on mailman3/hyperkitty roll out based on our discussions at flock.
- We now have a mailman01.stg instance. It's setup to use yum-cron and apply security updates every night (we should confirm it's working as we want it to tho)
- We need mailman3 / hyperkitty rpms added to infra repo.
- We need to get mailman3/hyperkitty roles added to this instance.
- Next we need to test importing things. I can get copies of the existing lists Mailboxes to import with.
- Lots of testing. :) Writing SOPS/docs for list admins, creating new lists, etc.
- Once testing looks good, we can move to production. Due to needing a database backend I was thinking we would just leave it in phx2 (and make 2 of them). smtp-mm hosts can easily be used for incoming emails and bastion for outgoing.
- We already have a few folks eager to move their lists over and help once we get to production. We can move this list (infrastructure) and cloud is very interested too.
- Once those initial lists look ok, we can schedule a flag day and move all the rest of the lists.
- Down the road, once RHEL7 is out we can migrate over to that.
Some questions:
- Are we going to need to keep the old archives online for people searching history/google? Or is there any way to redirect them?
- Should we do a seperate db for this in production? It's unclear to me how much in the way of resources the db side will take.
- Currently lists are at osu and don't go via proxies, but if we have it at phx2 we could just use the proxies and get caching, etc as well. Or is there a reason not to?
Any other things we should look at for the roll out?
kevin
On Fri, Aug 23, 2013 at 12:28:21PM -0600, Kevin Fenzi wrote:
- Are we going to need to keep the old archives online for people searching history/google? Or is there any way to redirect them?
I think we do need to keep them available somewhere, yes.
Auhhh....
Will Mailman RPM become a separate one 'mailman3' or just bump the version of 'mailman' package?
On Sat, 24 Aug 2013 15:24:31 +0800 Christopher Meng cickumqt@gmail.com wrote:
Auhhh....
Will Mailman RPM become a separate one 'mailman3' or just bump the version of 'mailman' package?
We discussed this some at flock. Short term it will likely be a seperate rpm. Once there's an official release and things are working we should/will look at updating the one in Fedora to version 3 most likely.
kevin
- We now have a mailman01.stg instance. It's setup to use yum-cron and apply security updates every night (we should confirm it's working as we want it to tho)
It doesn't look like I can access it, I probably need to go through lockbox, which I haven't needed to do before, so I probably don't have the rights, and I don't know the procedure. Could you point me to a SOP describing what lockbox is used for, what I can do with it, etc? I didn't find it in the SOPs wiki page. Also, how can I get to the private ansible repo and the host inventory? I'll make the configuration for this server in ansible.
- We need mailman3 / hyperkitty rpms added to infra repo.
I need to rebuild the HK rpms and change the MM3 RPM to make it not replace the current mailman RPM we have.
- Next we need to test importing things. I can get copies of the existing lists Mailboxes to import with.
Yes please! If you don't want to do the copying please just point me to their current location.
- Are we going to need to keep the old archives online for people searching history/google? Or is there any way to redirect them?
There is a way to redirect the old URL scheme to the new one but it's far from perfect (it uses the email import order which can be very flawed with duplicate emails). I'd rather keep the old ones.
- Should we do a seperate db for this in production? It's unclear to me how much in the way of resources the db side will take.
I admit it's still rather unclear to me. I am storing a full copy of each message in the DB, but I'm not sure it's even mandatory (it's used very rarely) and if it's a problem I could add a config switch to prefer a filesystem-based storage (maildir).
- Currently lists are at osu and don't go via proxies, but if we have it at phx2 we could just use the proxies and get caching, etc as well. Or is there a reason not to?
I don't see what problems caching could pose.
Happy to see this moving forward! Cheers!
Aurélien
On Mon, 26 Aug 2013 17:43:44 +0200 Aurélien Bompard gauret@free.fr wrote:
- We now have a mailman01.stg instance. It's setup to use yum-cron
and apply security updates every night (we should confirm it's working as we want it to tho)
It doesn't look like I can access it, I probably need to go through lockbox, which I haven't needed to do before, so I probably don't have the rights, and I don't know the procedure. Could you point me to a SOP describing what lockbox is used for, what I can do with it, etc? I didn't find it in the SOPs wiki page. Also, how can I get to the private ansible repo and the host inventory? I'll make the configuration for this server in ansible.
Catch me on IRC and we can sort out what you need. ;)
- We need mailman3 / hyperkitty rpms added to infra repo.
I need to rebuild the HK rpms and change the MM3 RPM to make it not replace the current mailman RPM we have.
ok.
- Next we need to test importing things. I can get copies of the existing lists Mailboxes to import with.
Yes please! If you don't want to do the copying please just point me to their current location.
Once you have access to the new instance, you should also have access to the existing mailman instance, so I can tell you where the files are there. Or I can copy them somewhere for you, etc.
- Are we going to need to keep the old archives online for people searching history/google? Or is there any way to redirect them?
There is a way to redirect the old URL scheme to the new one but it's far from perfect (it uses the email import order which can be very flawed with duplicate emails). I'd rather keep the old ones.
ok. It's a pretty big pile of stuff, but aside from disk I bet it won't take much resources.
- Should we do a seperate db for this in production? It's unclear
to me how much in the way of resources the db side will take.
I admit it's still rather unclear to me. I am storing a full copy of each message in the DB, but I'm not sure it's even mandatory (it's used very rarely) and if it's a problem I could add a config switch to prefer a filesystem-based storage (maildir).
Yeah, we may well have to just wait and see how it looks after we do some imports. If the stg db has problems thats a good indicator. :)
- Currently lists are at osu and don't go via proxies, but if we
have it at phx2 we could just use the proxies and get caching, etc as well. Or is there a reason not to?
I don't see what problems caching could pose.
Happy to see this moving forward! Cheers!
Me too. ;) Lets keep it moving along.
kevin
infrastructure@lists.fedoraproject.org
-
Aurelien Bompard -
Christopher Meng -
Kevin Fenzi -
Matthew Miller