On 11-12-01 12:19 PM, Zac Bentley wrote:
> We recently had an issue with a long Slony (2.0) catch-up after 10 days
> of disconnect between a master and a slave. It's possible that it was
> caused by either bug 167 or bug 222. To test this behavior we made a
> dummy cluster on our local LAN, and did the following steps:
>
> 1. Initialize a two-node Slony cluster with two identical copies of the
> database.
> 2. Allow initial subscription to catch up.
> 3. On the slave: drop the network connection to the master.
> 4. On the master: run ~4 million update operations.
> 5. On the slave: restore the connection to the master.
>
> We did this, and I was able to watch sync events get submitted and
> received in the logs. However, in sl_status, st_lag_num_events and
> st_lag_time kept going up, and the backlogged changes were not
> propagated (after a couple of hours, at least). The LAN link between the
> two nodes is fast, and neither node is lagging due to server/IO/network
> load. Why is this occurring/what did I do wrong?


Your 4 million row update was done as a single transaction.  This means 
that Slony needs to replicate it to the slave as part of a single SYNC. 
If slony is busy replicating that sync the SYNC won't show up as being 
replicated until that SYNC is done.  sl_status will show the slave as 
falling behind while the processing of that SYNC is going on (because it 
is behind, none of your 4 million rows will be visible on the slave 
until the transaction commits).

If you run the slon for your slave at debug4 you should see lots of 
DEBUG messages from the remoteWorkerThread showing progress.  Slony will 
fetch the rows in groups of 500 at a time with a cursor.  Based on how 
frequently you see these you can probably estimate how long it will take 
to do 4 million.



>
> Also, on what time period do you publish RPMs of Slony to public Yum
> repositories? (i.e. when should we expect to see an RPM of Slony 2.1?)
>
> Thanks!
>
>
>
> _______________________________________________
> Slony1-general mailing list
> Slony1-general at lists.slony.info
> http://lists.slony.info/mailman/listinfo/slony1-general