michael at aers.ca michael at aers.ca
Wed Dec 2 14:52:29 PST 2009
I've had something very odd happen. I'm not sure what caused the issue
so I'll try giving some background.

 

We have a cluster of 4 servers - 2 local (nodes 1 & 2) and 2 remote (3 &
4). Awhile ago we wanted to try a pseudo-test of our failover strategies
and so cut our VPN connection to the co-location and had it begin to
operate in failover mode. This meant the 2 servers here were continuing
to replicate in their cluster and the 2 remote were now operating in
their own cluster. Generally things went as planned, we learned some
things, and plotted out some changes.

 

I then removed the slony cluster information from the remote servers.

 

Today we brought the VPN link back up and I went about adding the remote
servers back into the cluster. We have a few different DBs, each with
their own clusters, and one of them has a very large table that takes
several hours to copy. To be efficient, after getting everything
listening right, I subscribed to all the sets *except* the one with the
very large table. I then made sure all the other clusters were working
properly. Things all seemed good.

 

So I came back and subscribed node 3 to set 8 (the one with the big
table), with node 1 as the provider.

 

Things looked fine. Slonik script executed, no errors, but then things
seemed to happen too quick - a matter of seconds instead of hours - so I
took a closer look. Everything says node 3 is subscribed to node 8, but
when I look at sl_table and sl_sequence on node 3, there's no tables or
sequences listed for node 8. The only thing I see in the log is:

 

DEBUG2 remoteWorkerThread_1: syncing set 8 with 0 table(s) from provider
1

 

Now, this set is replicating properly between node 1 and node 2. I have
verified that. So, what could cause it to end up like this on node 3? 

 

I have tried unsubscribing the node and re-subscribing but that didn't
help.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.slony.info/pipermail/slony1-general/attachments/20091202/156d6926/attachment.htm


More information about the Slony1-general mailing list