[Pgpool-general] pgpool-II dies when running a 'vacuum analyze'

[Bruno Lustosa]

>> The vacuum runs for some time, and then the pooler just dies. On the
>> client, I get the message about 'backend closing connection
>> unexpectedly'.
>
> Do you use VACUUM VERBOSE?

No, I didn't, at least not in the beginning.
In fact, I tried to run the vacuums in each backend separately, but it
didn't release the space. So, to solve the issue (and reclaim almost
20gb of space), I modified the code to stop inserting on the table,
and run a vacuum full on the table, in each backend. That released the
space, and now I can vacuum the table normally through the pooler.
When I tried to run the vacuum verbose (sorry, only tried it on the
backends, and not on the pooler), it complained about max_fsm_pages
being too low to be able to reclaim the space. There was also a "HINT"
message, but I'm not sure this is sent from backend (I think it's from
the client, but I could be wrong, of course).
Perhaps pgpool became confused about the "error" sent from the
backend, even though it's not fatal?
I can't reproduce it anymore, because the database is back to its
normal size, but as the vacuum would stop after a couple minutes, it
seems it stopped just when the vacuum would issue the messages about
not being able to reclaim the space.

Thanks!

-- 
Bruno Lustosa <bruno at lustosa.net>
http://www.lustosa.net/