Awhile back I tried clustering with R5 and wasnt very successful. It worked but the replicas would become out of sync and I had to clear replication history. The databases that we were using had between 100k and 250k documents. I am not a clustineg expert but it seemed pretty straight forward. Has anybody experienced this? Also looking for reback on clustering with R6 in general, sucess or horror stories.
Thanks
David
Subject: Clustering with large databases
David,
I haven’t had the chance to intensively observe R6 in a clustered heavy-use environment so far. However, in low-use environments and test environments, it wasn’t too bad, but I didn’t test it explicitly.
But I have one comment to your R5 experience: did you use a DEDICATED & fast network between the clustered servers? This is an URGENT recommendation not only by Lotus. If you cluster the servers while they are on the “normal” network everyone uses, delays may occur, and this could theoretically result in problems as you have described.
In general, when configuring a cluster, I use a dedicated Network interface card on every server which handles the clustering traffic.
Have fun,
Stephan
Subject: RE: Clustering with large databases
The cluster replicator does not rely on the replication history of the database, it maintains updates in memory and updates the history once per hour.
Normal scheduled replication would be coming along at the specified interval and re-sync the databases, this and not clearing the replication history was probably keeping you in synch.
As suggested in the other post, it is probably a cluster performance or configuration issue.
Subject: RE: Clustering with large databases
thank you for your response. Not sure my first question was clear. I currently have 2 servers with replicas of a large database. I usually replicate against one server, when I switch servers the client has to do a long search. This is painful when using a modem. I guess it is establishing replication history. Anyway I am hoping that clustering will allow the switching between cluster members without the long first replication. Does replicating against a cluster give this ability?
Thanks
David
Subject: RE: Clustering with large databases
Not in my experience. Each replica is unique concerning replication.