ADSM-L

Re: Client Session Stuck in MEDIAW Status During Migration

1998-03-18 12:55:23
Subject: Re: Client Session Stuck in MEDIAW Status During Migration
From: "Rushton, John D" <john.d.rushton AT LMCO DOT COM>
Date: Wed, 18 Mar 1998 12:55:23 -0500
Tom,
From what I have seen (and posted in my original reply to the Listserv),
the session goes to tape ONLY for the dataset that was forced to the
next pool in the storage hierarchy.  Once that file has been written to
that pool, ADSM will re-evaluate whether the next dataset can fit in the
disk storage pool, etc.

As I also mentioned earlier, if you have enough sessions in MediaW and a
limited amount of available tape drives, then this could take awhile.
What I have done in situations where this has happened at our site (and
it's happened more than once <G>) is to cancel any processes that are
using tape drives (e.g. tape reclamation - which takes 2 drives or even
migration itself if the disk storage pool utilization is low enought)
and even evaluate the sessions that are in MediaW - killing some of the
less important client sessions if they are ahead of the ones I care
about (the order that they are waiting in is reflected in the Media Wait
time reflected in the QUERY SESSION command.

I would definitely lower the high threshold below the 90% used by the
original poster (we use a hi of 80 and a low of 30 during the night -
changing it to a hi of 1 and lo of 0 at 9 am to force migration to tape
and then resetting it to 80/30 at night).   I would also review the tape
usage at nighttime to ensure that the migration process can get a tape
drive to offload the disk storage pool to tape.

John Rushton
Lockheed Martin
> ----------
> From:         Thomas A. La Porte[SMTP:tlaporte AT ANIM.DREAMWORKS DOT COM]
> Reply To:     ADSM: Dist Stor Manager
> Sent:         Wednesday, March 18, 1998 11:43
> To:   ADSM-L AT VM.MARIST DOT EDU
> Subject:      Re: Client Session Stuck in MEDIAW Status During
> Migration
>
> I've seen this situation occur with our disk pools as well. What
> seems to happen is a) the disk pool fills up (100% or very near
> to it), b) migration begins, c) sessions targeted for the filled
> disk pool now need to go to tape, b/c disk pool is full, so they
> go into MediaW state.
>
> Once (c) has happened, and the session has requested a tape,
> there seems to be no going back to the disk pool for that
> session, even if migration brings you back down to 0%
> utilization.
>
> I'd like to see this changed such that the session will
> reconsider the disk pool once migration has brought it down under
> the "High Migration %" value.
>
>  -- Tom
>
> "It's a dog eat dog world,                Thomas A. La Porte
>  and I'm wearing milkbone underwear."     DreamWorks SKG
>               - Norm Peterson
> <tlaporte AT anim.dreamworks DOT com>
>
>