cancel
Showing results for 
Search instead for 
Did you mean: 

Server Java Memory Usage

Former Member
0 Kudos

Dear,

On all 3 of our PI machines there is outrageous memory usage ...

The Dev machine has to be rebooted repeatedly because the memory usage shoots up to 2Gb, which is the maximum allowed ... Note that on the Dev machine are only running about 40 messages / day. The QAS machine does equally bad, though it has 4Gb memory, so it lasts longer.

The production machine only has a couple of 100 messages a day for now, as we're still only starting up with PI here, though the memory is already at about 2Gbs used ... What will happen if we're processing 60 different types of EDI messages each at a rate of 100's a day, and tens of thousands of messages to an SQL database per month?

We believe the SQL Adapter is the problem but we have no sure way of finding that out. We did note that if we don't send messages to the SQL database, the memory doesn't increase drastically.

The CC is set to close the connection after the message is sent.

Any ideas on what to do? We're not too enthousiastic about going live like this of course ...

Kind regards,

Frederik-Jan.

Accepted Solutions (1)

Accepted Solutions (1)

former_member200962
Active Contributor
0 Kudos

this can happen because of the logging/trace level settings....if you have used the maximum setting level then you are bound to fill the available space quickly.....to avoid this you can either decrease the level or turn off the logging (not recommended since logging for all the messages will be turned off)...maximum trace level is 3....min is 1 (which means trace is on)...0 means trace deactivated....

http://help.sap.com/saphelp_nw04/helpdata/en/d6/49543b1e49bc1fe10000000a114084/frameset.htm

/people/michal.krawczyk2/blog/2005/05/10/xi-i-cannot-see-some-of-my-messages-in-the-sxmbmoni

Check if the reply helps you in any way

Regards,

Abhishek.

Former Member
0 Kudos

The TRACE_LEVEL on the Development machine is 3, though it's only at 2 on the Production machine.

Our first idea was that indeed it's all the extra logging that was causing the quick buildup of memory used, but the extended logging (and there's a lot of it custom made) is only active in Develoment and on Quality Assurance, not yet on Production, and we're noticing unusually large amounts of memory used there too.

We have yet to try the way using the Solution Manager proposed here, but I'll be sure to keep you posted.

Answers (1)

Answers (1)

Former Member
0 Kudos

Dear Frederik,

Best solution is to analyze the PI system using Solman E2E Workload Analysis.

From your description I believe your j2ee engine has memory issue. Try to find answer for the following questions

1. Is my whole J2EE engine affected

2. Is the issue only with a particular server node?

3. Is it a H/W issue or J2EE application issue?

4. If its a H/W issue, does memory usage reduce after full GC?

5. What is the frequency of Full GC?

6. If the issue is with J2EE application, is it a memory leak?

7. If the issue is with the SQL Adapter , try to find the usage trend in Wily Introscope

Rgds,

Mat.