Memory Leak / Process Size

[Tom Rioux]

We have a process that keeps growing by about 2-3M everday.   The only was to keep this under control is to bounce the process.    However, I would like to get to the root of the problem.  We are on Cloverleaf 5.6 and here is what I have found so far:

1.  I’ve shut down the processes then checked the process log.  There are no active handles and there is nothing specified about a memory leak.

2.  I’ve checked the couple of xlates in the process but there is nothing notable in any of them that would cause an issue.  Most of the threads in the process are raw routes.

3.  I’ve checked the tcl procs that are being used by the various threads.  I can’t find any active handles.  All message handles are being dispositioned correctly.   I’ve even run them through the testing tool for verification.

My only question is this:

When the engine does it thing to check for active handles, does it check for any open file handles as “active handles”?   I did find one upoc that is being used in the tps read/write upoc protocol.   In the proc, I have found where a file is being opened for writing but can’t find where the file handle is ever closed.  I run this upoc through the tester and get ‘no active handles”.

Could this be the cause of my process continuing to grow?

Thanks in advance for any advice….

Tom Rioux

Topic

[Jim Kosloskey]

Tom,

I don’t believe the engine considers file handles in its handle checking.

This could very well be your culprit especially if the open is happening for every message rather than at start.

email: jim.kosloskey@jim-kosloskey.com 29+ years Cloverleaf, 59 years IT - old fart.

[Troy Morton]

Thomas,

Do you mean 2 to 3GB?  2 to 3MB for an hciengine process is not large at all.

I agree with Jim.  There is probably a leak somewhere, especially if you have very many Tcl Procs parsing the messages.

We also have one process that grows continuously and must be rebooted periodically.  There is a lot of Tcl in this interface and I’m sure there is a leak somewhere, but we haven’t found it.  Our process only grows about 20MB per day and we bounce it when it gets to 200MB.

Troy

[Author]

Replies