MHR wrote:
On Thu, Apr 1, 2010 at 9:25 AM, nate centos@linuxpowered.net wrote:
I *just* finished upgrading to CentOS 5.4 6 days ago.
How many people got trampled in the rush?
You might be surprised how many outages it takes to co-ordinate such an upgrade in a medium-large environment(and nobody including me likes to take *everything* down at once though we did have such an outage a few weeks ago to move a storage array I upgraded about 30 systems on that day). The fully redundant systems are easy to upgrade of course but there are lots of systems that are not fully redundant(and can't be made as such due to application design).
I tried doing some online upgrades for some of our more important systems(minus reboot for kernel) but something in the update wrecked havok on our NFS cluster the systems are very active doing NFS stuff 24/7. The NFS cluster recovered automatically but each time it took about 3 hours. I don't know what the upgrade might of restarted that would of impacted NFS activity. Since the upgrade there has been no repeats of the issue but during the upgrade within 30 minutes of upgrading active NFS clients(while they were doing stuff) caused immediate headaches on the cluster.
I suspect it's the first OS "upgrade" my company has done at least on linux. Looking through my inventory of systems these are getting a bit stale RHEL3/4:
1 AS release 3 (Taroon Update 3) 5 AS release 4 (Nahant Update 1) 6 AS release 4 (Nahant Update 3) 36 AS release 4 (Nahant Update 4) 1 AS release 4 (Nahant Update 6)
I don't count RHEL4->CentOS v5 as an upgrade since it is a complete re-install. For the most part those will get upgraded when the systems are retired I think.
nate