less than 10% of Virtual memory left!

[Eileen Pitts]

How would I go about figuring out why we get  “There is less than 10% of Virtual memory left!” message.  We are now getting this 1 – 2 times a week and it brings down processes.  We are on QDX 5.3 on a Unix box.

Topic

[Glenn Friedenreich]

Hi Eileen – Try lsps -s to see paging space used, e.g.

$ lsps -s

Total Paging Space   Percent Used

     2048MB              41%

You can then cycle each of your processes and repeat the lsps -s command after cycling each to see if the Percent Used decreases.  It sounds like you may have a memory leak issue.  We have this problem with two processes and need to cycle them every few days to release paging memory.  We’re on Cloverleaf 5.3.0 on AIX.  Hopefully when we upgrade we’ll see the memory leak problems go away.

– Glenn

[Eileen Pitts]

Thank you, I will do this.  How do we find out where the memory leak is so that we don’t have to cycle the process every couple of days?

[Glenn Friedenreich]

If there is a memory leak in a process, it is probably something that needs to be corrected by Quovadx (via version upgrade).  In our case, we’re running an older ODBC version that has this issue.

– Glenn

[Jim Kosloskey]

Eileen,

If there is memory leak in one of the Tcl procs used at your facility, you should see the handles which were not cleared in the log at shutdown.

Jim Kosloskey

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

[Eileen Pitts]

I looked at the log at shutdown and the to_sms_xlate found the leaked handles in the Xlate interpreter.  There are 3570 messages with under the Allocated by column.  Can you give me insight into where to go to find this mess and correct it?

[Russ Ross]

Thanks for your post because it caused me to look at my log file a little closer for some new TCL code I wrote.

Just so happens I discovered I had overlooked a memory leak because I called

grmcreate

and overlooked doing a corresponding

grmdestroy

to clean up the allocated handle before exiting my TCL proc.

On the hopes your memory leak could be similar, try searching your TCL procs for

grmcreate

and make sure if any occurrences of grmcreate are found that you make sure there is a corresponding call to

grmdestroy

to clean up the allocated handle before exiting the TCL proc

Here is an example of the extra line of code I ended up adding to solve my problem before exiting the TCL proc

if {[info exists ghd]} {grmdestroy $ghd}

For those that are curious here are lines from the log file that let me know I had a leaked handle

Russ Ross
RussRoss318@gmail.com

[Craig Weldy]

When you create a new script, you can test it and check for memory leaks by running it through the testing tool.  Just check the leak detection (don’t remember the exact name) button and run some test data through.  I have made a habit of checking this box anytime I am testing just to make sure I didn’t introduce a probem into my site.

This was only after doing just that (putlive leaking scripts) a couple times.

Craig Weldy
Senior Interface Analyst
Beacon Health System
South Bend, In, 46615

[Author]

Replies