[ZODB-Dev] ZEO server segfault under stress

[Tim Peters]

[John Belmonte]
> In ZODB-3.3a2 I'm getting ZEO server random segfaults quite frequently
> when running a little stress test of my high-write application.  The
> ZEO log isn't very informative, for example showing:
>
>      2004-03-01T18:35:24 pid 28737: terminated by SIGSEGV
>
> Even if I'm doing something relatively stupid in my application, I
> should expect an exception rather than a segfault, correct?

Yes.

> Checking the obvious like available memory and disk space reveals
> nothing interesting, so I guess I'll have to dig further.  If there are
> known issues with 3.3a2, please let me know.

I know of a thread race in 3.3a2 (recently fixed on the HEAD), but that's
it.  The worst consequence of that was rare test failures on Linux, and
more-frequent test failures on Windows, never a segfault.

Using a debug-build Python and debug builds of all other components is
usually the best thing to try first when facing a segfault.  I trust you're
already using Python 2.3.3.  I'd try it with ZODB HEAD, hoping it goes away
by magic then.