Amaya Gamarra wrote:
>
> Hi,
>
> I’ve got a 2 node’s slony 1.6.2.2 cluster and I’m inserting 400 
> records/second in tables with length 1 day. The records are very 
> simple so I don’t have any problems during normal synchronisation 
> processes. However, when I want to switchover the nodes, I find that 
> sometimes this operation takes me 1 minute and other times it takes me 
> 15 minutes.
>
> The reason is that the times it needs 15 minutes to achieve, the 
> cleanup thread is performing a delete logs operation that needs about 
> 14 minutes:
>
> DEBUG1 cleanupThread: 842.121 seconds for delete logs
>
> Can somebody tell me why this operation needs so much time and how can 
> I tune the application in order to reduce this delay?
>
> Thanks a lot,
>
> Amaya
>
Once every so often, Slony-I needs to delete old entries from the log 
tables, sl_log_1 and sl_log_2.

In version 1.2 and earlier, this takes place via DELETE, and if there is 
a lot of activity, then this will indeed take a not-inconsiderable 
amount of time.

In version 2.0, the DELETE is replaced with TRUNCATE, which should be 
rather faster.