Methodserver Deadlocks and alerting

[avillanueva]

We are now a week into Windchill 13 so lots of new things and experiences to understand but one that I do not remember seeing for a long time was deadlocks. I've seen two events were I had a deadlocked or dead Methodserver, both fixable by restart.

Aside from grepping logs for "Deadlock", are there any JMX alerts that can be configured to notify when this conditions arise? I thought this was a function of the ServerManager to detect and whack misbehaving methodservers and restart them.

Also monitoring SELinux from being the cause of all my grief.  I can see it blocking things randomly causing unexpected behaviour and threads to halt. Aside from that, any other places I should look to resolve root causes of deadlocks? 

2025-08-20 07:27:50,478 ERROR [wt.jmx.core.mbeans.Dumper.deadlockChecker] wt.jmx.notify.deadlock  - Time=2025-08-20 07:27:50.274 -0400,
Name=DeadlockNotifier,
SourceObjectName=com.ptc:wt.subsystem=Dumper,
class=class javax.management.Notification,
type=wt.jmx.core.mbeans.DumperMBean.deadlockDetected,
userData=[337,
37114,
37115],
message=Deadlock detected between threads: [337, 37114, 37115],
JVM Name=1112112@hostname,
Full Stack Trace=[Full thread stack trace (1112112@hostname; 2025-08-20 07:27:50.475 -0400),
      JVM: Amazon.com Inc., OpenJDK 64-Bit Server VM, 17.0.14 (17.0.14+7-LTS),
      OS: Linux, amd64, 5.14.0-284.30.1.el9_2.x86_64,
      Available processors: 8,
      Thread contention monitoring: disabled,
      Thread CPU time monitoring: enabled,
      Thread memory allocation monitoring: enabled,
 ,
Monitor Deadlocked Threads: 337, 37114, 37115,

[rhart · Accepted solution]

I've only just setup JMX notifications for the first time today, so I've no idea how common deadlocks are in my system.

I found the Mbean:

com.ptc > MethodServers > MethodServer.1234 > com.ptc > NotificationHandlers > DeadlockNotifier > Attributes.EmailRecipientListName.<JMX-Administrators>

looks like it's already setup to alert the JMX admins.

[avillanueva]

If you grep your logs for "Deadlock" you can see events. There is also a table in the database called JMXNotifications which seems to log events where it would have notified the administrator of something. I am already identifying my next tuning opportunity.

[rhart]

Does the DeadlockNotifier > Attributes.EmailSendLimit let you specify how many times a deadlock can be detected before the MS is reset automatically?

The default is 0 so I guess would just keep emailing forever with no reset.