P
Paul Koufalis
Guest
This advice should apply to 99% of OE environments: If you are on any version before 10.2B08, use the maximum allowable pica value in your version. In 10.2B08 the max was raised from 8192 to 1,000,000 (I think they were sick of raising it every couple of years). In that case, use 32,768. Larger pica values are required if a) the total AI writes is VERY high and/or b) the bandwidth to the target is relatively slow. You cannot base your pica calculation only on AI writes unless you have a very fast (read local) network. The minute you're going over a WAN, bandwidth varies wildly. Last week one of our customer's replication was falling behind every day and almost catching up every night. We were tracking it in ProTop, watching AI files fill and get locked and fill and get locked. Replication status was stuck in "synchronizing" mode and never quite got back to "Normal Processing". The network admins swore nothing was wrong and only 1% of the WAN bandwidth was being used. It must be a Progress bug!! Not. Copying a 1 MB file from source server to target server took minutes. Copying it back from target to server took one second. Yeah, must be a Progress issue. The lesson here is to monitor pica usage and to stop the replication server process before the pica buffers fill, otherwise forward processing for ALL users will be stalled and your phone will start ringing off the hook. Monitor replication server AND agent status (both sides) as there are bugs in earlier versions whereby the source will report "Normal Processing" and the target will report "Listening". ProTop does all of this and more. Regards, Paul
Continue reading...
Continue reading...