RE: HA failover: Nothing we try reduces client recovery below one minute

Hi, are you able to reproduce your issue without Kubernetes layer on Artemis instances? I’m not sure how exactly do you kill a pod, but 40 second timeout very much looks like default pod grace timeout + 10 seconds. -- Vilius From: John Lilley Sent: Wednesday, February 21, 2024 8:45 PM To:

Re: HA failover: Nothing we try reduces client recovery below one minute

Since your parameters didn't change the behavior, you could try tunning tcp settings (jdk and OS). I find the TCP stacks across different OSs behave differently. JDK example: -Dsun.net.client.defaultReadTimeout=2 -Dsun.net.client.defaultConnectTimeout=1 OS Example (windows): HKEY_LOC

Re: HA failover: Nothing we try reduces client recovery below one minute

I would recommend a couple of things to narrow down the issue... First, have a script or something on the client gather thread dumps at regular intervals (e.g. every 5 seconds) starting right before you kill the pod and continuing until the client fails over and recovers. Second, turn on TRACE lo

HA failover: Nothing we try reduces client recovery below one minute

Greetings! We are having a devil of a time trying to reduce delay during a failover event. We’ve set our URL to (tcp://dm-activemq-live-svc:61616,tcp://dm-activemq-backup-svc:61617)?ha=true&reconnectAttempts=200&initialConnectAttempts=200&clientFailureCheckPeriod=1&connectionTTL=1&callTi