Hi - in our location we are gradually moving into Tomcat/JBOSS application servers and away from Websphere. In the past we used the websphere agent with PMI enabled and monitored the web container thread pool size as a key alert metric.
In Tomcat, as we don't have PMI, we are looking at the "Threads|Active Threads" metrics. What we are seeing is that the value just grows and grows, it doesn't appear to reset until a server restart. So as a monitoring metric we aren't getting any value from it.
Any ideas/theories about why this metric value just grows upwards ? We think it may be that the thread objects are left on the heap and not gc'd, and the agent is returing that number as the "Active Threads" value.
We have tried to do a thread dump via the Investigator option but this is not allowed as the thread dump total stack trace exceeds 12,000 - is that a configurable value so we could increase it to get a thread dump in this way ?
In 10.1, we have very significantly upped the limit on the number of lines a Thread dump can contain (I believe it’s now 200.000). Prior to 10.1, you can up this value up to 25.000, this property controls it:
Configure the maximum stack elements the Thread dump can have,
If the user configures the max stack elements beyond 25000,
The property is reset to the default value of 12000