This is a post on how to search, analyze, display, and troubleshoot issues with Java platforms. JBoss products run on a Java Virtual Machine (JVM), and you’ll probably need to contend with memory management (MM).

Search/Locate the exceptions

If MM is not tuned properly, system administrators, developers, and other ops folks will encounter Out Of Memory Errors (OOME). But this could also happen if there’s an application bug. OOME can happen in different parts of the JVM. This post on JournalDev explains about the JVM memory model, and for its different parts there are different solutions click here to learn about details of the different kind of OOME.

This is a real world example from a customer where the memory wasn’t properly tuned. There were four EAP VM’s using SSO picketlink.

One of the nodes had "bigger" files in size and more in quantity. To immediately find them OOME, I execute this command:

cat server* | grep -i xceptions > xceptions.log

On the "exceptions file" created above, I can do more findings on the dates where these exceptions happened. Starting from here, if an OOME is actually found, one can do a deeper analysis.

Here is the example command:

strings xcepciones_TAM_JGO.log | grep -i outof | awk '{ print $1 " , java.lang.OutOfMemoryError "   }' | grep 2018 > oome_dates.csv

And here is the command output:

…
server.log.2018-06-21:1172:19:19:09,571 , java.lang.OutOfMemoryError 
server.log.2018-06-21:1188:19:19:41,482 , java.lang.OutOfMemoryError
server.log.2018-06-21:1200:19:20:27,231 , java.lang.OutOfMemoryError
server.log.2018-06-21:1201:19:20:31,713 , java.lang.OutOfMemoryError
server.log.2018-06-21:1202:19:20:33,365 , java.lang.OutOfMemoryError
server.log.2018-06-21:1229:19:20:35,813 , java.lang.OutOfMemoryError
…

In order to graph it:

Open the file with LibreOffice and separate as shown here:

Next, Insert a pivot table:

Select the row (Dates), column (Data) fields and which data to count (Data Fields).

Now, change the Data Fields to Count instead of Sum.

Show the same pivot table, but as a Graph.

Here are the results:

What does this reveal?

On these four days, there were at least two days where the server could not function properly and reported at least 100 times it had exhausted the heap. This tells us that proper memory management tuning must be implemented.

For this particular issue there was a misconfiguration with the default heap for server instances on the host.xml for EAP’s. For different types of OOME, different solutions must be provided.