We have also expirienced this problem on an R40 (AIX 4.1.4) running as an
NFS server.  We have not implemented HSM in a large scale yet, because of
this problem.  So we have limited number of users using the exported HSM
filesystems.  We have told those users to be *very* careful so they don't
get into "file system full" situations.  Additionally, you can set the HSM
parameters so you have better chance to have free space, premigrated files
or migration candidates enough to prevent the "file system full"
situations.  (Look at AUTOMIGNouse in your management class, and
-HThreshold, -Lthreshold and -Pmpercentage on the dsmmigfs Update command.)
nd.)

This is what hung up our machine: File system became full beacuse of
recalling a file that was bigger than free space while there were no
migration candidates.

Good luck!

Bjorn Harald Sandvik
Statoil

To:       ADSM-L @ VM.MARIST.EDU
cc:        (bcc: Bjørn Harald Sandvik)
From:     hthta @ SARA.NL @ INTERNET
Date:     05.11.97 11:20:37
Subject:  serious ADSM/HSM problems



Content-type: text/plain; charset=us-ascii


We have serious problems with ADSM/HSM on our SP2. Since a couple of
days our system hangs after recalling a couple of migrated files and
it seems it has something to do with (dead)locking nfs-filesystems on
which these recallprocesses are running. This blows up our fileservers
and after a while our whole SP2.
Our problem looks rather familiar with APAR IC15268 / PMR 24380 which
is announced two month ago. There is still no fix available.
My question is: does the site who came up with this problem has some
workaround so our machine keeps alive without disabling HSM?
(AIX 4.1.4, server 2.1.0.13, clients 2.1.0.7)
Thanks in advance,
Henk ten Have     SARA, Academic Computing Services Amsterdam
Btw: IBM is working on our problem.