Jeff Boes <jboes(at)nexcerpt(dot)com> writes:
> We had a production server crash last night. The postmaster log is given
> below:
> 2003-05-22 03:16:22 [26031] DEBUG: server process (pid 16661) was
> terminated by signal 14
It's really hard to see how that could happen. There is no point in the
lifespan of a backend where there is not a valid signal handler for
SIGALRM --- it inherits SIG_IGN from the postmaster, and changes that
to a normal handler routine before it ever enables the alarm anyway.
Color me baffled ... is it possible this is a kernel or glibc bug?
regards, tom lane