changing disks in LSM mirrored root partitions

From: John Speakman <speakman_at_biost.mskcc.org>
Date: Mon, 03 May 1999 17:55:24 -0400

Hi, sorry to swamp you guys with so many messages of late, but I just
discovered this list and it's at a time of dire need. I did run through the
archives and this is sort-of-but-not-quite covered:

we have a Trucluster (PS) cluster of two 4100s. Most of the disks are
on the shared SCSI bus except for /, /usr, /var, and swap, which as
Trucluster requires are mounted locally and then mirrored using LSM.

This question is in two parts, one lengthy and probably dumb, and another
brief and probably even dumber.

(1) We need to (why we do is a long story) remove one disk from all the two-disk
mirrorsets and replace them with new disks. For example, on one of the
systems rz8 and rz9 are a mirrorset, and rz10 and rz11 are another.
we need to remove rz8 and rz10 from this configuration and put new, unused disks
in their place, copy the labels and put the new ones back as mirrors. Does
anyone know how this is done? It is not straightforward owing to rootdgs and all
that stuff. Digital are umming and aahing a bit. This system is somewhat
critical
so we don't want more downtime than we can really afford and do not want to
find ourselves, as we so often have when tinkering with LSM, waiting with the
system unbootable at 2 a.m. in the computer room for support to call back
and help us out.

(2) We are tired of LSM and after all this time I'm ashamed to say we really don't

know what's going on. We need to mirror the system disks but they cannot be
on the shared SCSI bus. If we hardware mirror them using an HSZ or something,
are they still considered "local" by TruCluster and ths satisfy the requirement
for
a non-shared system disk? I can't think of any reason why not, but I'm wondering
why we didn't just do this in the first place three years ago and I can't remember

whether there was a reason (other than cost) not to just buy a local HSZ
controller
for each computer and stick the system disks on it. If it meant we could ditch
LSM
we would do it like a shot.

Thanks! I will summarize.
John Speakman, Memorial Sloan-Kettering Cancer Center, NYC
Received on Mon May 03 1999 - 21:58:07 NZST

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:39 NZDT