VxVM root mirroring resiliance

[David Nixon]

I have VxVM rootability set up, under HP-UX 11.11: L1000: two mirrored drives, on separate I/O channels.

I'm a bit disappointed by the behaviour resulting from a SCSI read error on rootdisk01 (my PRI boot disc).

After a retry count was exceeded, the path to
that disc was disabled. At this point many
processes were hung on I/O and the system
required a console re-set.

After rebooting from the ALT mirror disc
mirroring seems to have re-synchronised and
rootdisk01 again has ACTIVE status - though
,obviously, the underlying disc is now suspect.

My question is should VxVM mirroring of a root
disc seamlessly cope with disc failure, or do
the results depend on the amount of I/O going
on at the time?

[Steven E. Protter]

Shalom,

Your results seem to show the primary disk were going down while your system was running.

The fact a boot corrected this does not mean the disk is working now. You need to test it and most likely replace it or the component that caused the problem (cable, controller, drive cage).

If your system is properly configured the failure of the root disk should happen without you noticing it.

I suspect therefore the configuration is not correct. Once you fix the disk problem and re-mirror your system, do some testing to make sure the mirroring actually works.

SEP

[Coolmar]

Yes, the primary or alternate disk can go and you would not notice....other than alerts in the logs, etc. The failover would be seamless. It seems that there must be a problem in the configuration.

[Roland Piette]

Hi David,

Following my experience, you have to delete your /etc/lvmtab (or rename it would be better) before recreate it with vgscan -v.

Roland

[David Nixon]

The failure mode(s) of VxVM root mirroring
need further testing.