logical volume issue

[Asad Malik]

HI
Our system is a node in serviceGuard cluster. System rebooted with a panic. HP recommended a patch which has been applied. One funny Thing happened that in /dev/vgxxx dir, one of the logical volume lvxxx has time stamp of the time of reboot but the rlvxxx has the timestamp of when the logical volume was created.

Any insight why the time got changed and why only on lvxxx file.

Thanks

[Vincent Stedema]

Asad,

My guess is that most filesystem operations are done on the block device file. The character device file is modified only when the lvol is created and when you use newfs to create a new filesystem. Taking into account the fact that the system crashed, the modified timestamp probably has something to do with filesystem log replays.

Vincent

[Asad Malik]

HI
That means fsck was not run on this filesystem after the crash or is it something else, and is there a log file that may have an information that what happenned to this filesystem. Our system is also very slow since this crash.

[James R. Ferguson]

Hi Asad:

Interesting. I can't provide an answer, but:

1. I'm sure that this is not related to the panic nor to MC/ServiceGuard. I see the same modified timestamp *only* on /dev/vg00/lvol1 but not on rlvol1 on non-MCSVG servers that have been booted (shutdown gracefully) for normal maintenance reasons.

2. It's interesting that the filesystem for which the device file modification timestamp changes is an HFS one and is /stand (of course, since it is vg00/lvol1). Coincidently (or not?) the other filesystems (at least in my cases) are VxFS ones.

3. On my servers, the timestamp value actually matches the last time I started an Ignite make_tape_recovery. In fact, in one case, the maintenance boot I did afterwards did *not* affect the timestamp.

...JRF...