Home
7.2 Clustering FAQs
I don't really see how having a separate replicator for each domain can provide increased availability in the case where the other replicator goes down, since each of application instances would have a replicator defined and can't jump over to use the next replicator.
The assignment of application servers to replicators is to be viewed only as an initial startup mapping. If in the course of operations, one of the replicators in the domain goes down or becomes unavailable, the server will fail over to one of the remaining replicators in the domain.
If an application server goes down for thirty minutes, when it comes back up, would it receive all the invalidation requests from DRS? Or would it be out of sync with the rest of the application servers? If it is the first, is the only way out to disable flush to disk on startup?
When it comes back up, if flush to disk is off, it will be bootstrapped. If push/pull, it will only push metadata.
What happens if a replicator goes down? Would the second replicator pick up any invalidations that didn't get pushed by the first one?
Yes. Invalidations will be picked up automatically because both replicators see all invalidations.
Is there a recommended hardware guideline for DRS? One CPU per replicator? The size of the JVM?
512 MB is a reasonable size for the JVM. Updates are buffered in memory so memory is at more of a premium than the CPU.
What protocol does DRS use for communication in WebSphere Application Server 5.1?
JMS