Unicast Cluster Setup Is Not Working In RMB PERF Environment (Doc ID 2910818.1)

Last updated on DECEMBER 21, 2022

Applies to:

Symptoms

On RMB v2.9.0.1.0, Unicast Cluster setup is not working in the environment.

Unicast setting is not working on 2 node PERF batch server. It is a OCI VM with 8 core CPUs and 120GB memory. There are 2 Nodes per VM for batch servers. Both batch nodes are set up in unicast setting. TPWs have 4 threads each; so total 4 TPWs/16 threads expected to run in each node during a batch run.

We are getting error “Address already in Use…” when we start the second TPW in Node 2 after successfully starting all 4 TPWs in Node 1. So that means, we are able to start only maximum 5 TPWs/20 threads in both nodes together. Ideally we should be able to start 8 TPWS in both nodes (4 TPWS in node1 and 4 TPWs in node2). Also we find the issue is not specific to any node. If we start first 4 TPWs in Node 2 first and then we do TPW start in Node 1, the issue occurring for Node 1 as well when we start the second TPW there.

Team has tried following in tangsole-override.xml file:
IP1, port 1 and IP2, port 1 with auto-adjust-port = true
IP1, port 1 and IP2, port 2 with auto-adjust-port = true
IP1, port 1 and IP2, port 1 with auto-adjust-port = false
IP1, port 1 and IP2, port 2 with auto-adjust-port = false

We added auto adjustment feature in tangosol coherence file, but issue was not resolved. We find System is auto increasing port number for the TPW and possibly due to this, TPW using that port is unable to join the existing cluster.

Batch run has no error, but as we are unable to run the batches with all threads available. So it is causing a performance hit. Please advise what configuration/changes to be done for the affected TPWs and if any known issue for the current behavior.

Changes

Scenario_1:
Added port auto-adjust parameter "<port-auto-adjust system-property="tangosol.coherence.localport.adjust">false</port-auto-adjust>" as false and tried to start TPWs in Node1 and Node2.
- After one TPW successful start, second TPW not started with error 'SystemDatagramSocketProvider failed to bind to configured port 16481'.

Note: port auto-adjust feature has been explicitly disabled

Scenario_2:
Removed added port auto-adjust parameter or put value as 'True' and started TPWs on Node1 and Node2.
- 4 TPWs started on Node1 and 1 TPW on Node2.

Cause

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