> Gavin Hamill wrote:
>
>>I'll give a REINDEX a blast tonight as Andrew Sullivan's suggested -
>>hopefully that will reveal something strange, since we've been running
>>with Slony for a couple of weeks now fault-free - certainly longer than
>>the segfaults have been happening.
>>
>>
>
> Alas, a REINDEX of the affected table showed no output on the
> commandline or in the logs, and I'm still seeing fairly regular (roughly
> once a day) segfaults with exactly the same output from dmesg:
>
> postmaster[16589]: segfault at 0000000000000000 rip 00002aaaab5e8c00 rsp
> 00007fffffffd418 error 4
>
> As Jan Wieck was kind enough to help out picking through the core file
> that postgres dropped previously, (hence I knew what table was in use),
> I'm now back to square one with no idea why this crash is happening :(
>
> The most awkward part is that it causes the slon daemon to close, thus
> replication to all slaves is frozen until I get notified of the segfault
> and restart slon manually. :(

FYI, there is *something* of a way around the "causes everything else to
freeze" problem if it is a problem that the slon on the origin freezes.

There is a "gensync" script in the tools directory which you could run
every minute or so which simply generates a SYNC.

It is configured to only generate a SYNC if there hasn't been one in [some
interval of time] which means that if you pick the interval suitably, it
will normally do absolutely nothing.

That might at least help a bad situation...