TM1 Forum

Its just recently I tried to look into the }Stat cubes of TM1 9.5 (TM1 9.5 release in 2009...haven't tried using the newly release 9.5 last March)...I saw some threads of it not working properly. And I haven't tried it out in the previous versions of TM1 ever since the Cognos & IBM acquisition. Which amongst the version post acqusition have a good working stat cube.

Since the Latest being the only element in the }TimeInterval working...has anyone thought of a better way of doing the monitoring of the performance or activity being done in TM1?

appleglaze28 wrote:Since the Latest being the only element in the }TimeInterval working...has anyone thought of a better way of doing the monitoring of the performance or activity being done in TM1?

That would be very curious, you should see (open the relevant Subset editor):
LATEST
0M00
0M01
0M02
0M03
0M04
0M05
0M06
0M07
0M08
0M09
0M10
0M11
0M12
0M13
...

At least, this is the case in 9.4.1, and I doubt it would be different in other versions.

Wim

Hi Wim,

The point is that only the LATEST element is populated with data. In 9.4 and 9.5 none of the other members ever get populated with data with performance monitor on. This may be because the other members have been corrupted as they are created with ASCII character 4 proceeding the member name for the time based elements ... whether is a symptom, a cause, is totally unrelated coincidence, or has something to do with the Unicode rebuild who knows?

But the elements are actually defined in the dimension as:
0M00
0M01
0M02
0M03
... etc.

The answer to appleglaze's question is still a more or less straightforward on though - just build DIY performance logging cubes with whatever choice of time dimensions as appropriate and read regularly from LATEST in the }Stats cubes and write to the DIY cubes as per the system time.

lotsaram wrote:Hi Wim,

The point is that only the LATEST element is populated with data. In 9.4 and 9.5 none of the other members ever get populated with data with performance monitor on.

I can't speak to 9.4 FP / HF / SP / Whatever nomenclature is being adopted this week / 3, but that's not the case in 9.5 as I found from the test case that I submitted to Iboglix last week.

What i found was that the initial population of the elements was a bit random (perusing views in different cubes recorded only one of them) but that thereafter the correct }TimeIntervals elements were being populated, not just the "Latest" one. HOWEVER... it was only being done for the Admin user. (Not even users in the Admin group, but just the Admin user him/herself.) Anything that I did under the sa login wasn't tracked at all. Oh, and the Elapse Time (ms) element? It was rendering ridiculous values indicating that the calculation was taking between 11 and 22 years, but that's the least of the problem.

My suspicion is that the "quality control" relating to the supposed 9.4 "fix" was a guy logging in under Admin, just checking that some numbers, any numbers, appeared in the cube and saying "Yeah, that's fixed".

lotsaram wrote:The answer to appleglaze's question is still a more or less straightforward on though - just build DIY performance logging cubes with whatever choice of time dimensions as appropriate and read regularly from LATEST in the }Stats cubes and write to the DIY cubes as per the system time.

Won't fly under 9.5 if it's only logging one user.

The real solution is for Iboglix to pull their finger out and really fix this in 9.4 and 9.5.

lotsaram wrote:The answer to appleglaze's question is still a more or less straightforward on though - just build DIY performance logging cubes with whatever choice of time dimensions as appropriate and read regularly from LATEST in the }Stats cubes and write to the DIY cubes as per the system time.

I should have been more clear but I was only referring to server and cube stats. I have only been able to test on my XP laptop and possibly there are OS issues at play as in my case everytime I blow the }Stats cubes away and bounce the server (or create a brand new server for that matter) the StatsByCubeByClient cube is totally f'd as either the }Cube Functions dimension or the }StatsStatsByCubeByClient measure dimension seems to get corrupted so I end up with no working StatsByCubeByClient cube at all (so whether only the admin user gets updated is a moot point!)

Alan Kirk wrote:The real solution is for Iboglix to pull their finger out and really fix this in 9.4 and 9.5.

With you all the way, this is hardly a shining example of a QC process at work!