[Pgpool-general] saving state across pgpool restarts?

[Tomasz Chmielewski]

Jaume Sabater wrote:

> 2. After this event, pgsql in node 0 should be shut down by failover
> script configured in pgpool-II or any other mechanism.
> 
> 6. "For some reason" seems like kids playing around to me. I can only
> see this happening if all of the following are true:
> 
> a) No backup power available.
> b) Electricity goes down.
> c) No sysadmin receives an alert of what has happened
> 
> If you meant a software restart of the daemon, then you have a problem
> with humans, not with software.

Consider this likely situation:

- nodes are out of sync, pgsql is off on one of the machines,

- you need to restart both nodes (or, you're in the virtualized 
environment, you have no or little control over virtual host - when it's 
restarted, guests running on this virtual host will reboot as well).


Currently, there is no mechanism to make such restart safe.
If you're able to do it manually, you can restart the failed node, shut 
down pgsql, then restart the good node.


If you're not able to do restarts manually, you really risk data 
inconsistency.

After both pgsql nodes are brought up, pgpool-ii will just start talking 
to them without knowing that one of the nodes is out of sync. And later, 
finding the node which was out of sync would be real pain.



-- 
Tomasz Chmielewski
http://wpkg.org