"Eric B. Ridge" <ebr@tcdi.com> writes:
> Postgres locked up. All existing backends (roughly 100) couldn't execute commands. They'd just hang. One random
backendI selected had this backtrace:
> (gdb) bt
> #0 0xb7f7f410 in __kernel_vsyscall ()
> #1 0xb7e37a6b in semop () from /lib/libc.so.6
> #2 0x081916af in PGSemaphoreLock ()
> #3 0x081ba3c1 in LWLockAcquire ()
> #4 0x081b3b84 in ReceiveSharedInvalidMessages ()
> #5 0x0822d52b in AcceptInvalidationMessages ()
It looks like somehow the SInvalLock got stuck --- that would account
for both the stack traces you show. I'm not sure though why a "reload"
would appear to free things up. The SIGHUP signal for that would kick
the process off its semop() wait, but it ought to go right back to
waiting.
Have you tried restarting the whole database rather than halfway
measures like "reload"?
regards, tom lane
