This seems like my week for headaches.
A couple things here, the cause ultimatly is my fault but things just
don't seem like they are working as they should.
A tape stg pool ran out of scratch tapes, (oops, ouch)
First thing that doesn't seem right: Migrations run requesting same 2
tapes (2 migration processes) in filling status and receives End-of-Volume
message for each, then migrations would retry and do the same thing but at
2:20am it mounted the 2 tapes, got the end-of-vol message and dismounted
the tapes then sat for the rest of the night in a MediaW for those same 2
tape volumes, never mounting them and not in use anywhere that I could
tell.
The second thing that doesn't seem right: This tape storage pool is
collocated but the many sessions (over 24) that are in media wait are
waiting for the same 2 tapes that the migration was waiting for. Since
this is a collocated pool WHY do many different nodes want the same
tape?!? I would think that the session would be requesting the tapes that
the nodes have data on or at least a new scratch tape. I know that many
of these nodes still have space on their tapes. BTW we have 4 3590 drives
in the library that ADSM is using, and of all the media waits, no tapes
were mounted on any of them and not in use that I could tell.
Anyone care to elaborate if you can?
-*- -*- -*- -*- -*-
J. F. Holdren (` Unix Systems Group, Adv. Info. Tech.
jfh5 AT psu DOT edu PENN_)TATE Center for Academic Computing
(814) 865-2964 University Park, PA 16802
|