Thread

memory runaway on backend process

Creager, Robert S <creagrs@louisville.stortek.com> — 2001-04-02T16:28:05Z

Hey all,

I'm running 7.1RC1 on a Linux Mandrake system.  Another PC on the network is
running a Perl script which monitors my kernel log file and picks out
firewall hits.  It's been running fine for over a year.  I noticed this
weekend that the load on my PG server over 7.  I shut down the postmaster,
killed the backend processes hanging around, and tried to figure out what
happened.  I vacuumed the db, and fixed index errors.  I finally tracked the
problem to the backend servicing the network db client.  This client runs
through the kernel logfile and tries to insert all desired records,
expecting failures for duplicate entries.  Real simple stuff.  Near as I can
tell, when I upgraded to 7.1RC1 from 7.1beta5, the s**t hit the fan.  When I
launch the remote client, the backend consumes all physical memory (620Mb)
in 75 seconds, and starts swapping the server to a crawl.  When I stop the
remote script, thing go back to normal.  I don't even know where to start
looking for a problem.  Baring responses to the contrary, I'll re-install
7.1beta5 and see what happens, then try dumping/dropping/re-creating the db.
The other db's on the server are behaving normally.  I don't have the db
schema or scripts with me, and the main table has around 400,000 entries.

Thanks for any suggestions,
Rob

Robert Creager
Senior Software Engineer
Client Server Library
303.673.2365 V
303.661.5379 F
888.912.4458 P
StorageTek
INFORMATION made POWERFUL