[Pgpool-general] pgpool and Slony-I replication

[Jacques Caron]

Hi,

I am planning to use pgpool to handle switchover/failover between two DB 
servers replicating with Slony-I. I have a few questions, though, and I 
haven't found any information on those topics, don't hesitate to point me 
to the right place if I missed them.

The goal is to have two servers, without replication by pgpool, maybe with 
load balancing, and handle voluntary switchovers (during the nightly vacuum 
and other maintenances) as well as unexpected failovers.

Questions:
- in this scenario, pgpool will switch writing from primary to secondary by 
itself if the primary fails, right?
- likewise, pgpool will stop using the secondary (and send all reads and 
writes to the primary) if it fails?
- what exactly are the criteria for "failure"? No connection? No response 
to queries? What are the timeout thresholds?
- will pgpool forget about that host forever, or will it try to connect and 
reuse the server whenever it comes back?

- in some cases (during a switchover), reads can start from the new master 
right away, but writes need to be blocked for a few seconds. Is there a way 
to actually delay them? I.e. not send them to any of the servers, wait 
until some event happens (the switchover is finished) and then send them to 
the secondary? Reads should be done on the secondary right away, of course.

- how tested are the 2.5b1 "switch" features?
- do we agree that "pgpool switch" will stop sending requests to the 
primary (and thus send both reads and writes to the secondary) and "pgpool 
-s secondary switch" will stop sending requests to the secondary (and thus 
send both reads and writes to the primary)?
- how can one "restore" the primary or secondary without restarting pgpool 
(explicitly or implicitly)?

I guess that's it. Any help would be greatly appreciated.

Thanks,

Jacques.