Hi,
We have set-up a Service Bus 1.1 for Windows Server farm with 3 hosts and 6 message containers. In theory this should be a highly available set-up capable of withstanding the loss of one host. In our connection string we have configured each of the nodes (for example Endpoint=sb://<n1>,sb://<n2>,sb://<n3>;etc....).
To ensure high availability we have performed some testing and the results are not promising. In our test we have an application firing messages to a topic constantly. The service bus consistently receives messages at a decent rate with minimum latency. If we restart one of the servers or suddenly disconnect it from the network we typically loose a 1-2 messages due to a timeout (110 seconds). There after most messages will take anywhere between 3-20 seconds to submit to the topic. The servers don't appear to be under any load.
After the loss of a host I was expecting there to be an initial delay whilst the farm recovers (by moving message containers to another host). After the farm recovers I was expecting all messages to be submitted in a reasonable amount of time similar to the performance before the loss of a host.
Am I missing something or is intended behaviour? Cheers in advance for any input.
Thanks,
Craig