Installation fails

[BenLoosli]

I have installed Windchill 11.1 M020 CPS20 on 3 servers using the same downloaded files and a document I created when doing the first install to get the steps right. 

I have one VM server that refuses to complete the installation and aborts the install about 40% into the process after 2-1/2 hours of process the files. I have opened help desk calls with PTC on 2 prior fails on this same machine and they have been useless in figuring out why this one machine aborts the process. IT has looked at the VM machine and confirmed that it matches the other 5 Windchill/Oracle VM machines.

When it fails, I remove everything from both the WIndchill and Oracle servers including using regedit to remove all keys that point to the installation locations and then I start over. I am at a loss as to why this one machine aborts the process. The log files say it runs out of Java memory. But why just this one when I have done 4 other installs successfully ( I did one server twice as I verified my process.)

Has anyone seen this behavior from your servers? IT built a new set of servers 2 weeks ago and those installed first time through. I may clean these 2 machines off again and reverse the machines, putting Oracle on what has been the Windchill machine and Windchill on the Oracle machine. My machine names are s16pdmwind# and s16pdmora#. If I reverse them and Windchill installs, I will  use them as my test/development servers for Windchill 11.1 until I get new Server 2019 VM machines for Windchill 12.x testing at the end of the month.

[RandyJones]

I had this issue in the past: https://www.ptc.com/en/support/article/CS304896

[BenLoosli]

Randy,

I have admin rights on my Windchill servers. Is there a way that I can disable some of the cores, there are 12, so the install will complete? What would be a good number to reduce them to?

What I have a problem with is that I have 3 other servers, all with 12 cores, that have installed this build of Windchill successfully. It is just this one VM that is causing fits.

Ben

[RandyJones]

If you are experiencing the issue described in the article, you can reproduce it by simply installing WindchillDS by itself, as a test, which is an easy fast test. If that succeeds then you have some other issue if not then you could have this issue. I experienced this issue on 2 different Oracle Solaris boxes: 24 cores and 40 cores as seen by the os. Since I was installing into zones my solution was to temporarily limit the number of cpus in the zone, install Windchill, and then remove the cpu limit.

However my latest installs are on a RedHat Linux Dell server in which the os sees 32 cpus and I haven't had this issue.

I am going to guess with "only" 12 cpu's that your problem is not this particular issue.

[Cliff]

Ben

Will you post the PSI logs here?

- and please look the Windows Event Viewer around the time.

Cliff

[BenLoosli]

Cliff,

There is an open hep desk call and the log files have been uploaded to the call, C16240372.

It will take some digging to get to the Event Viewer logs and I am on vacation, be back in the office next Wednesday.

Ben

[Cliff]

Hi Ben,

The root is clear. Java process cannot start with heap of 2.6G because OS is short of memory.

- GWT (used by structure) compiling is resource consuming, then PTC configured the heap as -Xmx2560M in ant task

- While the OS is short of memory.

To fix the issue:

- Windows Event Viewer may report which process consumed OS ram. If other process consumed RAM, please uninstall/disable.

- Considering increasing OS RAM to e.g. 16GB; at least configure the Virtual Memory (Paging File) to be managed automatically.

- Considering reducing the memory of existing apps, e.g.

 1. Configure Oracle to below; if it is SQL Server, please configure accordingly via Studio (by default, SQL Server will use ALL RAM of machine).

Login Oracle via sqlplus with the DBA account, e.g sqlplus / as sysdba
SQL>alter system reset memory_max_target scope=spfile;
SQL>alter system reset memory_target scope=spfile;
SQL>alter system set sga_max_size=768m scope=spfile;
SQL>alter system set sga_target=768m scope=spfile;
SQL>alter system set pga_aggregate_target=256m scope=spfile;
SQL>shutdown immediate
SQL>startup

 2. Reduce the memory for Method Server

" The Error in installation log:

INFO 2022-01-31 18:57:46 - [java] Executing 'C:\Users\privbhl\AppData\Local\Temp\I1643664647\Windows\resource\jre\bin\java.exe' with arguments:
INFO 2022-01-31 18:57:46 - [java] '-server'
INFO 2022-01-31 18:57:46 - [java] '-Xmx2560M'
INFO 2022-01-31 18:57:46 - [java] '-Xss2048k'
INFO 2022-01-31 18:57:46 - [java] '-Dwtcodebase=E:\ptc\Windchill_111\Windchill\codebase'
INFO 2022-01-31 18:57:46 - [java] '-XX:CompileCommand=exclude,org/eclipse/core/internal/dtree/DataTreeNode,forwardDeltaWith'
INFO 2022-01-31 18:57:46 - [java] '-XX:CompileCommand=exclude,org/eclipse/jdt/internal/compiler/lookup/ParameterizedTypeBinding,<init>'
INFO 2022-01-31 18:57:46 - [java] '-XX:CompileCommand=exclude,org/eclipse/jdt/internal/compiler/lookup/ParameterizedMethodBinding,<init>'
INFO 2022-01-31 18:57:46 - [java] '-classpath'
INFO 2022-01-31 18:57:46 - [java] 'E:\ptc\Windchill_111\Windchill\codebase;E:\ptc\Windchill_111\Windchill\srclib\wnc\Associative-
INFO 2022-01-31 18:57:46 - [java] 'com.google.gwt.dev.Compiler'
INFO 2022-01-31 18:57:46 - [java] '-logLevel'
INFO 2022-01-31 18:57:46 - [java] 'INFO'
INFO 2022-01-31 18:57:46 - [java] '-war'
INFO 2022-01-31 18:57:46 - [java] 'E:\ptc\Windchill_111\Windchill\codebase/gwt'

...

INFO 2022-01-31 18:57:46 - [java] Error occurred during initialization of VM
INFO 2022-01-31 18:57:46 - [java] Could not reserve enough space for 2621440KB object heap "

Thanks,

Cliff

[BenLoosli]

My biggest question is WHY?

I have 3 identical virtual machine sets for Windchill and it installs on 2 sets but this third set fails. All the servers have 32GB of RAM installed. Why would the installer act differently on just this one set of servers. I also have another set of servers that are physical machines and it installed on them, too.

I will play with the Oracle commands when I get back in the office next week and then try to do the install again.

Should I strip everything off and start fresh or just resume the failed installation after changing the Oracle settings?

[Cliff]

"All the servers have 32GB of RAM installed" is big enough for new installation.

Regarding "WHY", it is not related with PSI Installer, just OS cannot launch a Process with 2.5G memory; two possible reasons  #1)something consumed RAM of the OS #2) OS problem (appears same as others, but might be not), configuring Virtual Memory/Page File big enough could help.

Do not need start refresh. Just look what we talked before.

Thanks,

Cliff