"Gaetano Mendola" <mendola@bigfoot.com> writes:
> Is it normal that after a while is like if the statistics collector fork a
> lot of time in only one second ?
> Oct 11 16:51:28 pserver2 postgres[8898]: [1943390] DEBUG: reaping dead
> processes
> Oct 11 16:51:28 pserver2 postgres[8898]: [1943391] DEBUG: statistics
> collector process (pid 16882) exited with exit code 1
> Oct 11 16:51:28 pserver2 postgres[8898]: [1943392] DEBUG: reaping dead
> processes
> Oct 11 16:51:28 pserver2 postgres[8898]: [1943393] DEBUG: statistics
> collector process (pid 16884) exited with exit code 1
> ...
No, this is not good.
Unfortunately, you don't seem to be getting any error message showing
*why* the stats collector is quitting. After a look at the code I note
that the 7.2 version of the stats collector writes fatal errors to
stderr always, rather than using elog() which can be redirected to
syslog. (This has been fixed for 7.3, but that'll do you no good
today.) You'll need to send the postmaster's stderr someplace other
than /dev/null so you can figure out what the problem is.
regards, tom lane