We just ran into this (or a situation very similar) last night on our
ADSM/6000 server. The problem was caused by ADSM putting 2 of our disk
storage volumes into read-only status because of an I/O error. Migration
would not trigger, and backups started going directly to tape. I lowered
the HI and LO thresholds to trigger the migration.
I guess a bunch of the files on the R/O storage volumes were cached files,
so the %MIGR was not high enough to trigger migration. On the other hand,
because the volumes were R/O, ADSM could not use these volumes to store
files on. I think a fix might be to calculate %MIGR and %UTIL using only
R/W storage volumes in the storage pool.
..Paul
--
On Fri, 11 Aug 1995 12:32:38 EDT Melinda Varian said:
On Fri, 11 Aug 1995 12:32:38 EDT Melinda Varian said:
>On Fri, 11 Aug 1995 at 09:19:49 PDT, Joe Faracchio said:
>
>> Could you double check that you were on .14 when this occured?
>> I am also curious why you want to turn off cache'ing. It seems to be
>> a good thing and can even help minimize tape mounts in a co-location
>> setting.
>
>Joe, I am actually running Level 14 with an additional fix for the
>loops in EXPORT NODE. I could never get any of the levels between
>11 and vanilla 14 stable enough for production here.
>
>I don't want to turn caching off. That was one of IBM's first
>recommendations for the problem we were having. I plan to turn
>it back on after SHARE.
>
>We still have 12% of that storage pool categorized as non-migratable,
>and I still have absolutely no idea why. Last night, with highmig
>set to 70%, utilization hit 100% again (during a migration), so I
>have lowered highmig to 60%.
>
>Melinda
|