Hi all,
we are using Postgresql 6.4.2 on FreeBSD 2.2.8 and have a lot of
problems with suddenly dying backend processes. We have already
changed kernel parameters to get more shared memory and start the
postmaster with -B 1024. The postmaster's virtual memory limit is
currently 300MByte and our largest tables contain about 12000
records. vacuum runs nightly and reports no errors. The problem
arises mostly after a couple of INSERT or SELECT INTO statements,
but also a 'COPY mytable FROM stdin' fails when I try to load
a file with 13000 datasets. It works when I split it up into
several files of not more than 5000 datasets each and load them
separately.
Turning on debugging for the postmaster as well as the backends
is not very helpful for us because the server seems to behave
differently. For Example, with debugging turned on we can't
create a certain view, which works without the debugging options.
Anyway, debug level 3 produced the following message:
postmaster: reaping dead processes...
postmaster: CleanupProc: pid 15608 exited with status 139
postmaster: CleanupProc: reinitializing shared memory and semaphores
shmem_exit(0) [#0]
binding ShmemCreate(key=52e389, size=8852184)
Does anyone know what 'status 139' means or where we can find
a hint in the source?
Any further ideas how we can track down the cause of our dying
backends?
Thanks in advance
Mirko