[Pgpool-general] Running pg_dump drops secondary backend

[Bruno Lustosa]

Hello.

I had a problem with previous versions. Whenever I tried to make a
dump of the database using the pooler instead of using one of the
backends directly, it would hang new connections, apparently because
of locking.
After installing from cvs, decided to try it again, after all, doing
the backup through the pooler should be better than relying on one of
the backends to be alive.
Backup did not stop, but the pooler disconnected the secondary
backend. This is what I got from the log:

2008-02-19 15:45:06 ERROR: pid 6277: pool_process_query: 1 th kind c
does not match with master connection kind d
2008-02-19 15:45:06 LOG:   pid 6277: notice_backend_error: 1 fail over
request from pid 6277
2008-02-19 15:45:06 LOG:   pid 22797: starting degeneration. shutdown
host db03.omicron(5432)
2008-02-19 15:45:06 LOG:   pid 22797: failover_handler: do not restart
pgpool. same master node 0 was selected
2008-02-19 15:45:06 LOG:   pid 22797: failover done. shutdown host
db03.omicron(5432)

I was monitoring the processes with 'ps ax', and saw nothing accumulating.
Any idea? I think pg_dump does 'COPY table name TO stdout', if I
recall correctly. I can turn on query logging and try to make it crash
again, if needed.

Thanks again!

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