On Thu, 13 Aug 2009, Les Mikesell wrote:
Geoff Galitz wrote:
...so I know for a fact updates can break a running system.
On CentOS? Fedora does that all the time but _not_ having behavior-changing updates in the long life of a major release is most of the point of 'enterprise' distributions. It's probably not perfect - and I wouldn't do auto-updates on production servers either but it should at least be very unusual for a CentOS update to break anything.
You mean like every time new Apache2 updates come along they _shouldn't_ break my failover cluster???
I've gotten into the habit of doing this each time I see an "httpd" update come down 'cuz a reboot will kill the cluster again:
If I did a reboot I need to edit haresources and restart heartbeat to get my DRBD drive back, then add links back in:
*) Remove "httpd" from the end of /etc/ha.d/haresources *) Restart heartbeat so that it'll mount my DRBD drive. *) Fix up the symlinks in /replicated/etc/httpd/:
logs -> ../../var/log/httpd modules -> /usr/lib64/httpd/modules run -> /var/run
*) Add "httpd" back into the haresources file. *) Restart heartbeat to check that the cluster comes up ok.
"httpd" updates remove the symlinks (that maybe "drbdlinks" creates? Can't recall) which point into my DRBD drive.
If I didn't do a reboot yet then I can fix things up with a subset of the above 'cuz the DRBD drive will still be mounted.
The above is probably necessary because I'm doing something wrong... Do I need to break the cluster before the update so that Apache2 doesn't touch the DRBD drive/drbdlinks stuff, then restart the cluster? That seems likely.