[pgpool-general: 356] DISALLOW_TO_FAILOVER ...

[Matt Wise]

We're using PGPool 3.1.2 in production right now with 4 DB servers (1 master, 3 streaming slaves). We're not quite ready to let PGPool handle automatic failover of our servers ... we're working on our scripts for this, but we just aren't there yet. To prevent PGPool from doing anything silly, we have all 4 of our nodes set with the DISALLOW_TO_FAILOVER option. If one of these nodes goes down, theres no action to fail it over. Great!

The problem is, if one node goes down, all access to the DBs is hung until that node is available again. This means that we basically have 4 single-points-of-failure rather than having a great additional amount of redundancy. I don't suppose I could request a patch that allows a single node to go down and traffic will just be redirected to another node until that one comes back online?

—Matt