8.98.2.3 Management kernel issues

ttrudeau

ttrudeau

Member
First post here, so bear with me...

My senior CNC and I have recently complete an install of 8.98.2.3. We're currently in a pre-development stage but have ran into a couple of issues. I was wondering if anyone has done a similar install and encountered these issues, or if anyone knows of a fix for them.

We are running our Enterprise server on an as400 V6R1. We're running JAS on Windows 2008 R2 with WebSphere 6.1. Our deployment server is also a Windows 2008 R2 server.

Our issue:

We've noticed after we start E1 that the Management kernel and SAW kernels are constantly increasing in memory utilization. The Management kernel also grows in CPU utilization until it is constantly using an entire CPU core. I'll attach screen shots from Server Manager and I'll attach our jde.ini.

We'd appreciate any help as we've been dealing with this issue since 8.98.2.0 and Oracle hasn't been able to come up with any resolutions or work-arounds.
 

Attachments

  • 155320-EntServer process detail resource data.PNG
    155320-EntServer process detail resource data.PNG
    99.7 KB · Views: 137
I've attached the jde.ini to this one. Oops on the first.
 

Attachments

  • 155321-jde.txt
    16.7 KB · Views: 294
Not sure how to post multiple attachments yet, but here's a graph of the CPU/Mem utilization.
 

Attachments

  • 155323-EntServer process detail resources.PNG
    155323-EntServer process detail resources.PNG
    111 KB · Views: 127
[ QUOTE ]
Do you see anthing in the JDE.LOG or the AS/400 joblog?

Tom

[/ QUOTE ]

We haven't been able to find anything abnormal in the logs as of yet. I meant to point out that we do not have any active users other than CNCs on the system at this point.
 
I have logged an SR with Oracle, yes. We've been working on it for quite a while. They've determined that the memory usage is being incorrectly reported to server manager, however we know the CPU usage is not. It's very easy to see how many CPU cycles have been used by the management kernel and it's pretty ugly.

We were seeing this issue on 8.98.2.2 as well, we upgraded to 8.98.2.3 to see if that would alleviate any of our issues. No joy on that though. If we find a working solution I'll post it here, and hopefully you'll do the same if you find one. Glad to hear we're not the only ones dealing with this.
 
Here's the Doc ID that explains the memory issues:

E1: KER: Kernel Resource Manager Running on iSeries Reporting Incorrect Memory Accounting (Doc ID 959703.1)
 
We ended up figuring out this issue. It was being caused by a couple of things. The Server Manager resources monitor turned out to be the culprit.

When we turned that monitor off, our the CPU usage for the Management kernel completely leveled off. I found that the default configuration 'Maximum number of samples to retain' is set at 1440 cycles. This by default holds 24 hours of data. This data has to be rebuilt by the Management kernel every 60 seconds to create the graphs for Server Manager. I decreased the 'Max retain' to 60 to store only an hour of graph data. This brought the CPU usage by the Management kernel to an acceptable level for us.

Secondly, I found that the resource monitor was causing the JVM garbage collection to run almost constantly. I changed the type of garbage collection from the defaults to 'GENCON' (generational concurrent). I also allowed 512Mb to be used as the object nursery. This dropped the amount of garbage collections that took place, while also speeding up the actual clean up time. You can find more info on the different garbage collections types here: http://www.ibm.com/developerworks/java/library/j-ibmjava2/

Let me know if anyone has any questions. I know I'm not the best at explaining all this.
 

Attachments

  • 157048-JVM settings.PNG
    157048-JVM settings.PNG
    37.7 KB · Views: 104
Back
Top