Folks,

We are running HSM (version 3, Release 1, Level 0.6L, a fixtest), on
an SP2 node (AIX4.2.1) with 1GB memory. We support one large
HSM filesystem, and currently have about 2.4TB of migrated data
in the form of 632467 migrated files. The dsmreconcile process
typically runs for 4 hrs while it traverses the HSM filesystem, before
creating a candidates file.

Over the past couple of days, dsmreconcile has been running for
about 4 hrs, and before completing, dumps core, never building
a candidates file. According to the AIX errpt, the process is
catching signal 4, SIGILL, illegal instruction. I have tried
running dsmreconcile from the command line and redirecting stderr
and stdout to a log file, but nothing gets logged when the process
dies, it just bombs.

Interestingly, if I run dsmautomig (and the candidates file is empty),
the process runs to completion, eventually generating a candidates
file. It was my understanding that under these circumstances,
dsmreconcile is called by dsmautomig. I don't understand why
the behavior is so different.

Before I call this in to ADSM Support, thought I would ask if
anyone else has run into this situation.

Thanks for your input.

--Ruth
  Cornell Theory Center/NCS
  Cornell Theory Center/NCS