I'm using pgpool3.0.4, and I'm not 100% sure if this scenario is ok.
I have two postgreses in replication mode, so all the queries are sent
to both backends (write-only, of course). When one node fails, and I
want to put it back to pool, here is what I do:
- shut down postgres on failed node
- rsync all the data from the failed node
- shut down postgres on running node
- rsync again
- fire up both postgreses
After that I need to restart pgpool, and then issue pcp_ command to
attach failed node. However, if my application is constantly connected
to the pgpool, there is a period before I attach failed node while all
the write-queries will go to only running node, and my databases will
not be in sync any more.
So, what I do is, I shut down pgpool, change the port it is run on, fire
it up, pcp_attach_node, than shutdown pgpool again, change the port back
to 9999, and fire it up.
Is there a way to tell pgpool NOT to allow any connections to both
backends (just to queue them)?
Or is there a better way to perform what I am trying to accomplish here?
Mario
_______________________________________________
Pgpool-general mailing list
[email protected]
http://pgfoundry.org/mailman/listinfo/pgpool-general
