Since few days we see that the Integration Server ( running on Linux ) is utilizing high memory…it is set with max 2GB java heap…in the same environment we have another identical Integration Server ( on another Linux virtual server )with the same configuration and custom packages and that is working fine.
when the memory utilization goes high even with not much heavy traffic…the swap process on the linux comes into picture and also the io% becomes abnormal and it doesn’t come down.
If we kill the integration server process and then suddenly the io% comes to zero and everything appears normal on the virtual machine. This indeed tells us that there is something fishy with the Integration Server and that’s the same argument that the Infrastructure team ( Unix admin ) say.
Did check and done everything possible to see if any particular package or service is causing trouble or anything else in the config parameters. Was unable to identify anything. As mentioned in the opening para, the other identical integration server with same version, fix level, deployments is working perfectly normal.
Did anybody notice such issue ? or can help me with this.
Also If your environment deals lot of packages/interfaces and volumes of transactions you need to adjust the min/max memory to the same size (for example) in your server.sh or setenv.sh (8.x) and restart the IS.
MIN_MEM=5000M
MAX_MEM=5000M
Make sure your OS has enough memory before you bump up the sizes above.
No It is not Prod…it is one of our test environment. I cannot bump it to 5000M as the total memory allocated to the machine is just 4096M…Done multiple restarts and even it takes lot of time to stand up…definetly IS is not getting enough memory to load all packages and save the config files.
As part of my investigation, I just noticed something…when I killed the IS process ( the only wM process running on the machine ) and did a top on the linux machine, to my surprise I could see that the memory utilized on the machine was still 72% . Ideally there should be nothing else on this machine apart of wM IS process that should run. So , going by this there is something that is utilizing the RAM. Just posted this info with screenshots to the infra guys, let me see what they come up with.
Plz ask your Unix Admin team to trace the memory utilization @ different intervals of time to findout the exact culprit. Plz post the updates to the forum members.
Yes…until now they were in denial mode saying that the wM java process is one utilizing the high memory…but we have asked them to come up with the metrics of memory utilization for each process…When we have set the java max memory to be 2G…they claim that out of 4G total physical memory of the VM, 2.7G is utilized by wM process…which I see it as absurd…
Some times there are lot of issues with JVM itself that can utilize memory and leaks etc…and sys admins gives tough time when comes to troubleshooting exercise on the environment and point the fingures in the other direction etc…it happens:)
But is there any reason why you are still on 8.0.1 and not on 8.2 (SP2) due to lot improvements done in the IS server wise? Have you think about going that route as well?
Also make sure with your admins that always 4gig memory available at any time on your loaded environment for better performance aspects (spikes) and see if they can help in any ways on the OS land scape across sharing applications.
There is an actual Linux server and a virtual Linux server, both running on identical IS. and seems TS facing High memory consumption on the Actual one…
It would still be strongly recommend to compare on 2 actual servers or 2 virtual servers. as memory allocation might not be same for both actual server and Virtual server.
If you want to compare them, Please make sure both having the same packages. With same package is being ENABLED/DISABLED at both server
