We are getting ready to upgrade to 6.2R1 on our F840C's and I was wondering what procedures some of you follow when doing upgrades. I am especially interested in hearing from those with Oracle databases NFS mounted to UNIX servers.
Do you shutdown the databases before doing anything?
I think this would be prudent. However, we aren't storing Oracle databases on our filers, so someone else would have more experience.
Do you leave everything NFS mounted running as normal during the entire procedure?
You can usually get away with this. Of course the NFS clients will get "NFS server not responding" errors during the downtime. But if the filers come back up soon enough, everything usually recovers.
We store user home directories on our filers and we have dozens of NFS clients in several different buildings, so it is not convenient for us to shut them all down.
Do you leave clustering enabled and just takeover-giveback to complete the reboot after the download?
You definitely do not want to takeover/giveback. Clustering is not designed to allow you to do software upgrades with zero downtime. It would be nice if you could have filer A takeover, upgrade filer B, then have B takeover and upgrade filer A. But this doesn't work because clustering doesn't work unless both filers are running the same release.
You should "cf disable" clustering before booting the new release so that each filer boots the new release independently of the other. Once they are both up, "cf enable".
You want to do this because there is often a disk firmware upgrade that runs automatically the first time the new release boots. If clustering is enabled, then each filer gets upset when it sees that its partner has disks with down rev firmware. This happened to us on our last upgrade, and things worked out OK, but we got a lot of disturbing errors on the colsoles, and both filers sent several autosupport emails.
If the new release upgrades your disk firmware, then the filers will be down longer than for a normal reboot, depending on how many disks you have. We have a clustered pair of F820s, each with about 25 disks, and the last time we upgraded, the disk FW upgrades added about 15 minutes to the downtime.
Any pros/cons and horror stories are greatly appreciated,
Steve Losen scl@virginia.edu phone: 434-924-0640
University of Virginia ITC Unix Support