E9.2 JDE Web Client Performance Slow Since Upgrading to 9.2.6.3

[mansat1]

Since the Tools upgrade, logging in to JDE and opening pages has been many times slower than what it was prior the the upgrade. Additionally, the slowness shows up intermittently during the day - no pattern that we have seen. No backup or other maintenance process involved.

Has anyone encountered this? We have a case open but it is waiting for Development team feedback. So posting it here too.
We came across a possible performance fix in the web pack 9.2.6.4. But installing that has not helped our situation.

Configuration:

iSeries V7R3 Enterprise Server

WebSphere 8.5.5 FP 21

 

[Holmes]

Was this ever sorted out? I am experiencing the exact issue.

 

[DrWhat?]

We've run into this too after moving from 32x to x64 9.2.6.4

Along with a host of other performance issues... slow postings, user disconnects/session timeouts...

 

[CHo]

We've run into this too after moving from 32x to x64 9.2.6.4

Along with a host of other performance issues... slow postings, user disconnects/session timeouts...

HI.

Which platform are you on? I can't see your configuration signature.

 

[mansat1]

This hasnt been sorted yet. I have a Sev 2 case open with Oracle for last 4 months! Sitting with Development team without any action.
@Holmes - what is your configuration? Same as mine?

 

[Holmes]

Thats correct @mansat.

WAS 9.0.5.10
iSeries v7R3
Tools 9.2.6.3 from 9.2.2.5. Also applied web pack 9.2.6.4 with no improvements.

 

[jolly]

Did this not show up at all in testing the tools upgrade?

Anything unusual in the JAS logs?

 

[MFreitag]

Did this not show up at all in testing the tools upgrade?

Valid question.
...but nevertheless interesting that some people seem to have run into the "same" problem.

For everybody that has this problem ( @Holmes @mansat1 @DrWhat? ): Let's try to find the similarities here. Are all of you running Websphere and is said Websphere running on your IBMi installation? If so, try to spin up a Oracle Linux Install with Weblogic or Websphere as HTML/AIS/BSSV... Server and see if it behaves the same.

 

[Holmes]

Our DEV ENT server does not have great resources as our Prod Box. Slow perfomamce during testing was somehow acceptable.

Our web servers are running on windows server 2019.

There is nothing unusal on the web server or webshpere logs

upon scratching around, Iv noticed the ASP on the Iseries box is above 80% full, which i suspect could also contribute to the slow perfomamnce

 

[jolly]

It’s been a while, but my recollection is that the iSeries isn’t even getting warmed up until ASP is well over 90% so don’t think that’s it. I’m wondering about the ClientAccess driver versions and config, 32/64 bit compatibility etc.

 

[Tom_Davidson]

You may want to put the JAS.log into debug, and look for long gaps. I was able to show on any earlier release that I had a 5 second gap in a function call/return (I've forgotten which one), using that Oracle was able to figure out the issue. It took a long time to find the gap though. If you've had a Sev two open that long, it is probably worth your time to comb through it (yes line by line).

JMHO.

Tom

 

[MFreitag]

Our DEV ENT server does not have great resources as our Prod Box. Slow perfomamce during testing was somehow acceptable.

Our web servers are running on windows server 2019.

There is nothing unusal on the web server or webshpere logs

upon scratching around, Iv noticed the ASP on the Iseries box is above 80% full, which i suspect could also contribute to the slow perfomamnce

We have two customers which we're moving off of Windows Server 2019 now to Oracle Linux 8. For some reason that nobody was able to explain the TCP stack on 2019 is going bonkers on some setups. It's worth a try at least and doesn't take too much time.

 

[mansat1]

We have two customers which we're moving off of Windows Server 2019 now to Oracle Linux 8. For some reason that nobody was able to explain the TCP stack on 2019 is going bonkers on some setups. It's worth a try at least and doesn't take too much time.

Interesting. I have never explored the Oracle Linux VM installation. Let me investigate and see if I can implement a test box.

 

[mansat1]

One question I have for those experiencing the slowness - Do you see this problem intermittently?
I have noticed that there are times when the system performs at acceptable speeds. But not able to identify the reasoning or any correlation to events etc.

 

[Holmes]

Our current workaround is to restart the ENT services which is not feasable. The problem goes away then comes back the next day.

 

[HolderAndrew]

In our experience when upgrading to 9.2 (TR 9.2.5.3) things in general were slower - especially when first opening applications on web client. Turned out to be a few things. Check your firewall to ensure everything is whitelisted. Also check your roles - is the experience worst for specific users? We had some users with many hundreds of different rules which were check and consolidated. I agree with above comments inspecting JAS logs (and others) to find the time gaps and focus on this.

Hope this helps.

 

[NewCNCGuy]

We had something very similar when we did our 9.2.5.x tools upgrade. I tracked it down to the JAVA memory settings in Websphere console. The typical path is something like "https://servername.domain.com:9043/ibm/console". They got reset back to defaults when we patched WebSphere as part of the upgrade. You can find the settings if you login to your WebSphere in Servers/Server Types/Websphere Application Servers, then chose one of your JAS instances. From there you get the config settings. To the right side you should see "Server Infrastructure" and under that Java and Process Management. If you click the "+" you should see "Process Definition". This will open up yet another screen, and to the right you should see "Java Virtual Machine". You should see Initial Heap Size and Maximum Heap Size settings here. We are on 64-bit, and we have ours set to 1024MB/4096MB in production.

 

[MFreitag]

We had something very similar when we did our 9.2.5.x tools upgrade. I tracked it down to the JAVA memory settings in Websphere console. The typical path is something like "https://servername.domain.com:9043/ibm/console". They got reset back to defaults when we patched WebSphere as part of the upgrade. You can find the settings if you login to your WebSphere in Servers/Server Types/Websphere Application Servers, then chose one of your JAS instances. From there you get the config settings. To the right side you should see "Server Infrastructure" and under that Java and Process Management. If you click the "+" you should see "Process Definition". This will open up yet another screen, and to the right you should see "Java Virtual Machine". You should see Initial Heap Size and Maximum Heap Size settings here. We are on 64-bit, and we have ours set to 1024MB/4096MB in production.

Interesting - do you remember what patch level you were coming from and going to? That shouldn't happen...

 

[NewCNCGuy]

Interesting - do you remember what patch level you were coming from and going to? That shouldn't happen...

I think it was the CNC who updated the WebShere that reset it, not part of the patch. He might have even did a full uninstall reinstall with the latest version (it has been a minute).

 

[Holmes]

I just wanted to update everyone.

We logged a call with IBM and they could not assist resolve the issue.

WE then decided to install weblogic 14c and the perfomance is lightning fast.