Very Computer

We are experiencing an odd problem with a database load  which we are
wondering has been encountered before by anybody out there ...

Details:

        Progress v 6.3F01
        AT&T Unix SVR4 2.3.1
        AT&T 3550 2 x i486 50Mhz
        1 Gb Memory

The database is a 2.5 Gig Multi-Volume DB and the load is running as a series
of 6 background sessions running concurrently loading from ascii dumps on
disk. Session 1 is loading all files sequentially except for a very large
transaction file,  the remaining 5 are each loading a section of this
transaction file. We also have 2 APW's 1 BIW and a WDOG running and the server
is started with no-integrity.

The load proceeds well for 30-40 minutes then *all* activity stops entirely.
The processes still exist (sleeping) and the clients show up in promon. The
activity screen shows that absolutely nothing is happening at all even after
several minutes - a few checkpoints had occurred earlier but nothing as
drastic as this.

sar showed that when the db goes to sleep the machine is twiddling it's thumbs
i.e. no swapping, cpus' 98% idle, no wait for i/o, low usr/sys time, the disks
are effectively doing nothing as well .... no errors on the console and
everything else looking and acting normal BTW we are the only users's on this
box.
The database log file shows nothing untoward either.

Looking at the promon User Control screen showed that all self service clients
were waiting on a TXE transaction commit lock except 1 which was waiting for
BKEX exclusive lock on a db buffer ...

Has anyone any ideas on why this 'blockage' appeared after 30-40 mins and how
it can be cleared or prevented ....? We saved the promon snapshots if anyone
needs more information.

All advice/tips/pointers are gratefully received.

Richard Gray
===========================================================
AT&T GIS European Deployment Centre
Sophia-Antipolis, France
VP:     327.4560
Tel:    +33 92.94.45.60
Fax:    +33 92.94.45.45

===========================================================
When all you have is a hammer, everything looks like a nail
===========================================================

I run on a Vax, not unix, but progress is progress (almost). If your
DB servers have been up for a period of time, you may have accumulated
what I call phantom users. These are users that have disapeared from
the operating system,  but are still registered as logged into the
progress database. A phantom user can have locks still applied to
certain records, and will stop the load dead in its tracks. Shutting
down the servers and restarting them is the cleanest way out of this
problem.

If this is not the case, you may have a corrupted index on that file.
Progress says tat loading massive amounts of data will go faster if
you disable the indexes on the files you are loading, then rebuild the
indexes after the load. This woul also eliminate any possible index
corruption problems.

Also, the BI file is an indexed file itself. I have had situations
where the database would lock up because of a bad index in the BI
file. Of course, truncating the BI file would eliminate that.

You may already be aware of all the above, and if I am stating the
obvious, my humble apologies.

Jacksonville, FL.
Where JESUS is Lord!