cplv weirdness

[stevieray]

I am attempting to split up a large vg and copying (oracle raw as it happens) lvs from sourcevg to destvg1 and destvg2.
When I try to run two cplvs in parallel, even though the destinations are different and the source disks are different, cplv seems to take a lock on the source VG and I have to do the copies one at a time.
This seems weird and unnecessary. Anyone got any good ideas why this should be so?

Steve.

 

[PSD]

Steve,

In AIX whenever you use/interrogate a VG is puts a lock on the volume group. This is for obvious reasons, I mean lets say you have two people both chaging VG's at the same time, one persons change a millisecond before the other guy presses return could change the state of the VG completely, and the change you made may not even be relevant. I mean bear in mind that the ODM database needs to be protected from corruption.

If you want you can use the chvg -u (vgname) command to remove the lock, but I would not recommend this in that I have never tested what happens when you remove the lock and perform another VG locking operation.

I am unaware of another easy around this.

I hope that helps

Paul

 

[ElgisRamon]

Hey! Think about this:

You issue a cplv and the LVM reads the VGDA from the disks and starts copying the volume, then you unlock the volume and start copying.... it reads an uncongruent VGDA and starts copying. Now the first command finishes... it will write what it thinks it must write. Then, the second command finishes and it writes what it thinks is the new VGDA: the original (before the first cplv) plus its changes [red]without[/red] the firts cplv's changes!

These commands must be serialized! [sig][/sig]