WebSphere Scaling

We have daily Hung Threads in WebSphere Logs and sometimes it causes extreme slowness and makes it impossible to login Portal. We have 1 Application Server and support has said we need 3. Has anyone else had this problem and did verticially scaling resolve the issue?

Find more posts tagged with

Infor Lawson Technology Group - Discussion

Comments

unknown

We are on LSF 10 and WAS 7.0.0.25. Some of our slowness was resolved by going from one server to three in a cluster as support mentioned. It may be good to check WAS's Java heap settings (KB 1206644) and ask support if they are set properly for your configuration. Whether you are on AIX, Unix or Windows there may be other considerations to review, i.e. our users who upgradred to IE11 have better application response times. Good luck.

unknown

Yes, we had a similar problem some time ago. One application server wasn't enough. We have been running with a cluster of three application servers for quite some time with no issues.

strahoon

We used have the same issue and we were scaled because lawson said that would resolve our hung job issue. It did not. We migrated to new servers several years after our original go live, we did not scale, and our system worked much better.

unknown

We believe it is the traffic from the Add-ins that is causing the hung threads. We recently had a merger adding 6,000 + new employees to the system, I think the added stress of data migration maybe causing this to happen more frequently. We are updating WebSphere this weekend to 8.5 and see if we see a difference. It's hard to duplicate this in a test system without heavy traffic. For those that are using 3 App Servers on AIX how much RAM are you using?

bhillz

You are likely running into resource issues by running large Add-Ins updates.

We have a policy that Add-in uploads are limited to 5000 entries. This helps spread the load and allows the system (app server and backend) to recover.

gwhiteford

We did several things to improve our performance. First, we clustered all the web servers except the one for Addins, we kept that separate so that it could handle large queries or addins. We also adjusted our JVM Min/Max Heap settings. Since then we have had no issues with queries or Addins timing out.

unknown

Thank you it is good to know. Do you know what your JVM Min/Max Heap settings are set to? Lawson is also recommending to map AddIns on a separate Application Server and Port.

gwhiteford

I can go look them up for you. However, our settings may not work for you. This has a lot to do with sizing. Your chart could be twice the size of ours and you could have double the amount of Accounting Units than we do, not to mention we don't have that many users, not compared to other Lawson clients I've been to. Infor should be able to give you some hints on this. They helped us with our heap settings. I would try opening a ticket with the Tech group and give them some of your sizing information. But, first I would see what happens when you cluster your servers and keep Addins out of the cluster and on its own.

If you are still having time out issues with Addins or queries after that, you may need to boost the RAM on your server. We did that as well.

unknown

Yes our Min or Initial Heap 512 and Max 2048
We have nearly 7000 back office users and 55,000 emss users. This system houses Finance, GL, HR, PR and Supply Chain

unknown

Hi Kathy, with 7000 back office users and 55,000 emss users, would you mind to share the CPU and memory information of your application server? We are doing load testing of a new deployment and it appears we run into a resource limit. Therefore it will be helpful to get some idea from your organization's configuration. Thank you!

0911060553270289.pdf

unknown

We are on AIX 7 across Production
LSF = 6 CPU 58 GB RAM
DB2 = 5 CPU 110 GB RAM
Landmark = 2 CPU 48 GB RAM (USED for ProcessFLow only)

unknown

@CGWare, How do you limit uploads from Add-in's to 5000? Is it just a local company business practice? Or is there another software block you implement?

unknown

There are only about 200 users that have ADDins

bhillz

Institutional policy. I wish there was a way to limit. We notice it when the user's ignore the policy. There's no hard stop at 5000, We don't quibble at 5001 or 5043, but at 12000 the impact is felt. Unfortunately, the user is in control of how large their Add-In sheet. They can make it as big as they want -- or think they can get away with.

Unfortunately, IT recommendations don't mean as much as one complaint from their own colleagues. Explaining it as: Having 5000 users hitting the Enter key at the same time. Once they realize that, (and a few complaints from their colleagues) things usually run better.

bhillz

@Kathy Williams: We found that Heap sizes have more impact than the number of CPUs attached (or even I/O capability).

Red Gold use of Role Security in LX.pptx

unknown

So do you increase or decrease the heap? I would imagine if we create 3 App Servers scaling in a cluster we will need more RAM to accomodate the heap for each App Server right?

moellerg

You could set one of your extra app servers up specifically for AddIns. There is no real way to force the issue, but if you have an app server (port) specifically for AddIns, then that app server gets the heavy hit and it's not mixed in with your normal user traffic.

We don't have this yet set up here, but are thinking about it.

We have a 4th app server (port) on top of our 3 app servers for users specifically for MSCM, though, and it appears to help.

dEPM-679 - 2020-11 Whats new.pptx

bhillz

Heap sizes are generally increased, but you may also need to adjust the garbage collection also.

Definitely test on your non-production servers first. Just be very careful about what you tweak. And always YMMV - Your mileage may vary.

unknown

We are running WAS 8.5.5.6 and 10.0.6 applications and 10.0.7 technology. I will am to add an Add-in server today in test and move to production next week. We are an employer of 2000 and we have two application servers currently. We have already adjusted the heap size. We are taking the same steps to address similar issues but we do not have near the user community.

unknown

@Greg How do you direct activity (ie AddIns or MSCM) to a specific server?

Copyright © 2025 Infor. All rights reserved.