Diagnostics Workload Analysis - Java Memory Usage gives BI query input

Question

Dears
I have set up diagnostics (aka root cause analysis) at a customer side and I'm bumping into the problem that on the Java Memory Usage tab in Workload analyis the BI query input overview is given
Sol Man 7.0 EHP1 SPS20 (ST component SP19)
Wily Introscope 8.2.3.5
Introscope Agent 8.2.3.5
Diagnostics Agent 7.20
When I click on the check button there I get the following:
Value "JAVA MEMORY USAGE" for variable "E2E Metric Type Variable" is invalid
I already checked multiple SAP Notes like the implementation of the latest EWA EA WA xml file for the Sol Man stack version.
I already reactivated BI content using report CCMS_BI_SETUP_E2E and it gave no errors.
The content is getting filled in Wily Introscope, extractors on Solution Manager are running and capturing records (>0).
Did anyone come accross this issue already?
ERROR MESSAGE:
Diagnosis
Characteristic value "JAVA MEMORY USAGE" is not valid for variable E2E Metric Type Variable.
Procedure
Enter a valid value for the characteristic. The value help, for example, provides you with suggestions. If no information is available here, then perhaps no characteristic values exist for the characteristic.
If the variable for 0DATE or 0CALDAY has been created and is being used as a key date for a hierarchy, check whether the hierarchies used are valid for this characteristic. The same is valid for variables that refer to the hierarchy version.
  Notification Number BRAIN 643 
Kind regards
Tom
Edited by: Tom Cenens on Mar 10, 2011 2:30 PM

Answer 1

Hello Paul
I checked the guide earlier on today. I also asked someone with more BI knowledge to take a look with me but it seems the root cause analysis data fetching isn't really the same as what is normally done in BI with BI cubes so it's hard to determine why the data fetch is not working properly.
The extractors are running fine, I couldn't find any more errors in the diagnostics agent log files (in debug mode) and I don't find other errors for the SAP system.
I tried reactivating the BI content but it seems to be fine (no errors). I reran the managed system setup which also works.
One of the problems I did notice is the fact that the managed SAP systems are half virtualized. They aren't completely virtualized (no seperate ip address) but they are using virtual hostnames which also causes issues with Root Cause Analysis as I cannot install only one agent because I cannot assign it to the managed systems and when I install one agent per SAP system I have the message that there are already agents reporting to the Enterprise Manager residing on the same host. I don't know if this could influence the data extractor. I doubt it because in Wily the data is being fetched fine.
The only thing that it not working at the moment is the workload analysis - java memory analysis tab. It holds the Key Performance Indicators for the J2EE engine (garbage collection %). I can see them in Wily Introscope where they are available and fine.
When I looked at the infocubes together with a BI team member, it seemed the infocube for daily stats on performance was getting filled properly (through RSA1) but the infocube for hourly stats wasn't getting filled properly. This is also visible in the workload analysis, data from yesterday displays fine in workload analysis overview for example but data from an hour ago doesn't.
I do have to state the Solution Manager doesn't match the prerequisites (post processing notes are not present after SP-stack update, SLD content is not up to date) but I could not push through those changes within a short timeframe as the Solution Manager is also used for other scenarios and it would be too disruptive at this moment.
If I can't fix it I will have to explain to the customer why some parts are not working and request them to handle the missing items so the prerequisites are met.
One of the notes I found described a similar issue and noted it could be caused due to an old XML file structure so I updated the XML file to the latest version.
The SAPOscol also throwed errors in the beginning strange enough. I had the Host Agent installed and updated and the SAPOscol service was running properly through the Host Agent as a service. The diagnostics agent tries to start SAPOscol in /usr/sap/<SID>/SMDA<instance number>/exe which does not hold the SAPOscol executable. I suppose it's a bug from SAP? After copying the SAPOscol from the Host Agent to the location of the SMD Agent the error disappeared. Instead the agent tries to start SAPOscol but then notices SAPOscol is already running and writes in the log that SAPOscol is already running properly and a startup is not neccesary.
To me it comes down the point where I have little faith in the scenario if the Solution Manager and the managed SAP systems are not maintained and up to date 100%. I could open a customer message but the first advice will be to patch the Solution Manager and meet the prerequisites.
Another pain point is the fact that if the managed SAP systems are not 100% correct in transaction SMSY it also causes heaps of issues. Changing the SAP system there isn't a fast operation as it can be included in numerous logical components, projects, scenario's (CHARM) and it causes disruption to daily work.
All in all I have mixed feelings about the implementation, I want to deliver a fully working scenario but it's near impossible due to the fact that the prerequisites are not met. I hope the customer will still be happy with what is delivered.
I sure do hope some of these issues are handled in Solution Manager 7.1. I will certainly mail my concerns to the development team and hope they can handle some or all of them.
Kind regards
Tom