Okay, this is really easy since they added the mirrorvg command, but here’s the checklist anyway:
If you have several busy JFS filesystems all using one log volume, there is contention for the JFS log; which slows down all the filesystems. You can either add another log to some of the filesystems, or switch to JFS2, which can put the log inside each filesystem. To add another log:
Splitvg is a really cool feature of AIX 5L, but there is one problem. Each “copy”, must be on a different set of disks. Mirrored copies of each LV are numbered 1, 2, or 3 (1 for the first copy, 2 for the second, and 3 for the third). Splitvg has to be told which “copy” to split and recreate as a new VG. So, this means that all of the #1 copies must be on the same disks, and all of the #2 copies must be on different disks. Splitvg isn’t smart enough to work around this. So, there’s a quick little script to run on each PV in the volume group, if it’s all okay there won’t be any output:
If you think you may have contention for a filesystem, lv, or volume; you can use filemon to pinpoint the problem:
Clearing PVID’s from vpaths
If you want to move a volume from one RS/6000 to another, you should clear the PVID. Earlier versions of the SDD had problems if the disk had a PVID when cfgmgr was run. This will correct that problem:
chdev -a pv=clear -l vpath0
Creating a PVID for a disk
Sometimes you want to force a particular PVID on a volume. There is a command to correct this:
/usr/sbin/create_pvid disk PVID
Import an existing PVID from a disk
On an ESS, early versions of Subsystem Device Driver flakes out if your hdisks have PVIDs. You can get hdisks with PVID’s lots of ways, but it’s easy to clear out. Here’s a quick and dirty hack to clean them up. I know it’s dumb, but it works, and at 2am, it’s easy to know what it’s doing.
On a multi-node HACMP cluster without enhanced concurrent VGs, anytime you add a LV to a volume group, you have to make sure the other nodes see the LV. This will also fix other VG out of sync issues. You can either take everything down and do an importvg on all the nodes, or you can do a “Lazy Update”:
Sometimes, especially in a HACMP environment, one system will leave a bogus reservation on a PV. While this reservation is set, no other node can use that disk. I’ve also seen this when one machine crashes hard. When you try to use the disk, you’ll get a PVMISSING status on the offending disks. This used to be a real problem, but I haven’t seen it on newer OS and driver levels.