Q&A About Latency On The Use Of Automatic Failover With Database Replication HA Setup (Doc ID 2369224.1)

Last updated on FEBRUARY 14, 2024

Applies to:

Goal

We have a Messaging Server deployment on an HA setup across two datacenters (~6ms latency between the two datacenters). Each of these sites has 3 Messaging Server Stores configured with Automatic Failover with Database Replication.

Q1:

We need to perform a disaster recovery test across the two datacenters. The latency between the two locations is approx. 6 ms. Under normal circumstances, two proxies (MMPs) and three Message Store (MS) systems run on location 1. During the test, a failover is performed from all systems on location 1 to location 2, one by one. At the end of the test all five systems run on location 2. During the test there is some time (this can range from a few minutes to a few hours) that one or two of the three message store servers run on location 1 and the other message store server(s) run on location 2. How sensitive is the Berkeley Database replication to this type of scenario? During peak hours we see some 80 msgs/sec inbound, delivery to message store.

Q2:

How sensitive is the NFS storage when, for example, the message store would be running on location 1 and the NFS storage is running on location 2?
In general:
- Is there a rule of thumb regarding distance between datacenters or maximum latency for the use of Messaging Server Automatic Failover with Database Replication?
- What would be the recommended setup for this kind of configuration?

Solution

My Oracle Support provides customers with access to over a million knowledge articles and a vibrant support community of peers and Oracle experts.