Re: Linux/PostgreSQL scalability issue - problem with 8 cores

Jakub Ouhrabka wrote:

> We've tried several times to get stacktrace from some of the running
> backends during spikes, we got always this:
>
> 0x00002b005d00a9a9 in semop () from /lib/libc.so.6
> #0  0x00002b005d00a9a9 in semop () from /lib/libc.so.6
> #1  0x000000000054fe53 in PGSemaphoreLock (sema=0x2b00a04e5090,
> interruptOK=0 '\0') at pg_sema.c:411
> #2  0x0000000000575d95 in LWLockAcquire (lockid=SInvalLock,
> mode=LW_EXCLUSIVE) at lwlock.c:455
> #3  0x000000000056fbfe in ReceiveSharedInvalidMessages
> (invalFunction=0x5e9a30 <LocalExecuteInvalidationMessage>,
>    resetFunction=0x5e9df0 <InvalidateSystemCaches>) at sinval.c:159
> #4  0x0000000000463505 in StartTransactionCommand () at xact.c:1439
> #5  0x000000000056fa4b in ProcessCatchupEvent () at sinval.c:347
> #6  0x000000000056fb20 in CatchupInterruptHandler
> (postgres_signal_arg=<value optimized out>) at sinval.c:221
> #7  0x00002b005cf6f110 in killpg () from /lib/libc.so.6
> #8  0x0000000000000000 in ?? ()

Perhaps it would make sense to try to take the "fast path" in
SIDelExpiredDataEntries with only a shared lock rather than exclusive.

--
Alvaro Herrera                                http://www.CommandPrompt.com/
The PostgreSQL Company - Command Prompt, Inc.