Archive for 18th January 2010

Creating a stable Arch server.

This is how I prefer to deal with stabilty issues for my Arch Server, ymmv, and you may do it differently.
I prefer not to rewrite/redo the entire system just to create a stable server.

1. Mirror Current Repo to a Local Repo
2. Use local Repo on test servers
3. If no issues, push local Repo to Stable Repo
4. Upgrade servers from Stable Repo
5. Pull Current Repo to local repo and start over with testing again.

————————–
| Current Internet Repos |
————————–
|
|
———————-
| Local Repo-Testing | —– Test Servers
———————-
|
|
——————–
| New Repo – Stable | –>>->>- Stable Servers
——————–

I haven’t ignored any packages for my setups, I upgrade kernels and do all upgrades, I just test them locally on 3 test server machines before I do my remote production server upgrades.
Has worked pretty well for me. You could ignore kernel upgrades I suppose and make the thing even more painless to do, but for me the whole point of running Arch is to be as close to the newest releases on all software. I don’t have time to create my own packages, and I never saw the need to duplicate the stuff that has already been done. For me, creating my own “stable” Arch server, has been 99% testing before updating the production server. So basically I create my own stable server repo through testing.

My setup works for me, but I’m really only tracking/working with (sshd openntpd mysqld httpd postfix proftpd) for the most part. As long as those are stable and issues dealt with/figured out, then the production server is usually happy. Archlinux.me downtime has been 100% hardware related, and not due to any problems with Arch itself. A very brief downtime while converting to php 5.3, but that was pretty minor.
I honestly haven’t looked at kernel26-lts ……. probably should I suppose, but the regular kernel has always worked pretty well.

Anyway, just thought I’d share how I do this for myself.