Sam <sam@palo-verde.us> writes:
> A particular web application I am working has a staging version
> running one a vps, and a production version running on another vps.
> They both get about the same usage, but the production version keeps
> crashing and has to be re-started daily for the last couple days. The
> log file at the time of crash looks like this:
> LOG: could not accept new connection: Cannot allocate memory
> LOG: select() failed in postmaster: Cannot allocate memory
This looks like a system-level memory shortage. You might find useful
information in the kernel log. I'd suggest enabling timestamps in the
PG log (see log_line_prefix) so that you can correlate events in the
two log files.
regards, tom lane