Melvin Davidson wrote:

> Question on an anomaly.
>
> I am currently testing slony slony1-1.1.2 on 3 nodes configured as
> follows:
>
> node 1                    node 2                  node 3
> MASTER ------>  SLAVE1 -------> SLAVE2
>
> All works correctly.
> However, I noticed that after doing a switchover to SLAVE1:
> ==================================
> lock set (id = 1, origin = 1);
> wait for event (origin = 1, confirmed = 2);
> move set (id = 1, old origin = 1, new origin = 2);
> wait for event (origin = 1, confirmed = 2);
> ==================================
>
> slon on SLAVE2 stops.
> It can be restarted successfully, but I am wondering if this is
> expected behavior, as it is not documented that this will occur.
>
> Why would the other slave(s) stop during a switchover?

It *is* necessary, upon this action, that configuration be reloaded, so
that the various nodes are aware of the change of node roles.

In version 1.2, this probably won't involve the slon actually falling over.

Consider it "not a huge problem" that it does, at this point, fall over.

Historically, one of the things we would traditionally try first, if we
ran into a replication problem, was to stop and restart the slons.  It
would seldom cause significant problems, and often closing and reopening
database connections would resolve whatever was going wrong.

We don't do that quite so often; leaving slons up, and trying to figure
out what's up tends to be pretty useful in figuring out problem
conditions.  But if you're not planning to spend time tracing things
unless you're forced to, "restarting the slons" is often pretty effective...