We have a 4 node federation/AAC cluster and migrated to external Database (Db2) in our test environment and conducting load tests. Our Db2 is still kind of unstable at high load (more on that later) so we implemented IBM-recommended IVIA connection pool and timeout tuning parameters listed below. Unfortunately, these changes did not resolve the original stability issue, nor did they improve overall performance. Instead, we observed a clear latency regression during authentication flows.
Specifically, application login response times increased significantly, with delays reaching up to 4 seconds, whereas the expected and acceptable login time is under 1 second. This latency was consistently observed during the test window and was not present prior to applying these settings. Could any of these specific configurations be causing the latency issues? The application restored normal login response times immediately after the changes were reverted. Which setting below has the potential to cause such a delay?
The following configuration was in effect during the test:
-
isam_cluster.config.connectionManager.maxPoolSize = 50
-
isam_cluster.config.connectionManager.agedTimeout = 3600s
-
isam_cluster.config.connectionManager.connectionTimeout = 20s
-
isam_cluster.config.properties.blockingReadConnectionTimeout = 30s
-
isam_cluster.config.connectionManager.purgePolicy = FailingConnectionOnly
-
isam_cluster.config.connectionManager.enableSharingForDirectLookups = false
-
isam_cluster.hvdb.connectionManager.maxPoolSize = 200
-
isam_cluster.hvdb.connectionManager.minPoolSize = 0
-
isam_cluster.hvdb.connectionManager.agedTimeout = 3600s
-
isam_cluster.hvdb.connectionManager.connectionTimeout = 20s
-
isam_cluster.hvdb.connectionManager.maxIdleTime = 300s
-
isam_cluster.hvdb.connectionManager.reapTime = 180s
-
isam_cluster.hvdb.properties.blockingReadConnectionTimeout = 30s
-
isam_cluster.hvdb.connectionManager.purgePolicy = FailingConnectionOnly
-
isam_cluster.hvdb.connectionManager.enableSharingForDirectLookups = false
-
distributedMap.getRetryDelay = 500
-
distributedMap.getRetryLimit = 5
------------------------------
Rajkumar Godi
------------------------------