Andrew Sullivan wrote:
On Mon, Jun 18, 2007 at 08:18:59AM -0500, Jeff Amiel wrote:
I have 2 nodes subscribed to a single master.
One of them had some issues at one point, and while new events appear to
be replicating at the moment, there are somehow gaps in the dataset.
The situation you describe should be _impossible_. If it has
happened, we sure want to know why. Did you do anything else during
these issues? What were the issues anyway?
I would agree...but I am sure this is MY fault either through neglect
(old slony version slony1-1.1.5_1 ) or blatent mis-use (I was in a
particularly harried state at the time)
Back then this started, had an issue with the site hosting the second
subscriber node (couldn't access it) ....and every 30 minutes or so, the
master node slony would crash/restart because it couldn't communicate
with the subscriber node...which was no good (old version of slony
probably didn't help here)
I removed the node from replication (but I don't remember how...either
by dropping the node (via the "Replication' item in pgadmin) or some
other means) Once our remote site access was restored, I re-added the
node (via slon script to set up the subscriptions properly). I remember
this being only a few hours later. I apparently didn't check to see if
it was properly replicating.
On this part Friday, however, we noticed that it appeared that
replication to the DR site was not occurring at all (and probably hadn't
been since the incident)....and realized that the sl_node table had
no_active set to 'false' for that node. I then used a couple of slony
functions to re-enable the node....and replication begin again.
I know I made mistakes here....and having a concrete course of action to
take a node offline and bring it back online would be helpful. What IS
the proper way to temporarily drop a node and then re-enable it later?
Jeff
_______________________________________________
Slony1-general mailing list
[email protected]
http://lists.slony.info/mailman/listinfo/slony1-general
