Feeders and fed cells reported / logged

[Hippogriff]

Hi,

I continue my TM1 journey and have recently been looking at the Fishcakes stuff you can find at http://www-01.ibm.com/support/knowledge ... ers_N8010F and specifically trying to learn about basic Feeders.

While I'm going through this I read about MaximumCubeLoadThreads in the tm1s.cfg file and this allows me to see information at startup like:

3064 [] INFO 2016-02-22 15:30:11.159 TM1.Server Processing feeders for cube "purchase"
3064 [] INFO 2016-02-22 15:30:11.165 TM1.Server Feeders created in cube "purchase": new 3355, duplicate 0, ignored 0

Where I'm assuming 3355 is the number of fed cells... but when I open up the }StatsByCube Control Cube (Performance Monitor is on) and look at the LATEST time interval, the purchase Cube and Number of Fed Cells it clearly reports 0.

This is pretty consistent... although after deleting the Rule for the Cube, starting up the TM1 Server and re-creating the Rule and saving it, I did get it to show 3358, which is too close to not be reflecting the same thing.

Is it reasonable for me to expect the number reported by the "Feeders created in cube" log line and the number reported by the }StatsByCube Control Cube to be the same? Why on earth might the Control Cube be showing 0 pretty much all the time? I would love to get some insight into this so I can feel as though I trust what the }StatsByCube stuff is telling me (I do realise it updates every minute).

[Steve Rowe]

Just to check, have you go the stats monitor on when you are looking in the stats cube?

[Hippogriff]

Hi,

Thanks for replying.

I've set PerformanceMonitorOn=T in the tm1s.cfg file and when I right-click on the TM1 Server in Architect it offers me the option of Stop Performance Monitor, so I'm hoping I have what I need..? As I say, sometimes it has definitely reported a non-zero number for me, and other items appear to be reported fine too, like Number of Populated Numeric Cells and suchlike. I just haven't got much confidence in Number of Fed Cells being consisently reported properly... nor am I even sure it should match (or be close to) the number I see in the tm1server.log at startup, but that feels sensible to me?

Does that answer your question or am I missing something else?

[Steve Rowe]

Hi,
I just wanted rule out the obvious reason for the values being zero.

Honestly not sure why the result would be zero when the monitor is on. I'm wondering if the monitor is on when the server launches it behaves a bit different and the cube you want to analyse needs to have been looked at through the front-end first?

I doubt many people have the monitor on all the time in the wild due to the performance overhead, so I wouldn't be surprised if there were unknown bugs. Try removing the line in the cfg and then turning it on manually, do you get the "right" result then?

Cheers,

[Hippogriff]

I will try that later and report back. For the moment I'm told I need to do some real work, rather than 'playing around'... but I shall be back!

[Hippogriff]

OK,

So I closed Architect and I set PerformanceMonitorOn=F explicitly in the tm1s.cfg file - then restarted the TM1 Server, I still have MaximumCubeLoadThreads=3 set. As expected, in tm1server.log, I see:

14980 [] INFO 2016-02-24 15:11:36.092 TM1.Server Feeders created in cube "purchase": new 3355, duplicate 0, ignored 0

Good stuff.

In Architect, I firstly opened the }StatsByCube Control Cube and shimmey things around - getting all 0s for LATEST, }PerfCubes - purchase and }StatsStatsByCube. That's OK as the Performance Monitor is not on. I right click on the TM1 Server and start Performance Monitor, I recalculate the Cube but no change... I recalculate about 10 times (and maybe I go over a minute here) and then I see some numbers change - Number of Populated Numeric Cells becomes 8128, Total Memory Used changes etc..

But Number of Fed Cells remains 0.

I open up the purchase Cube and recalculate. I go back to the }StatsByCube and recalculate - nothing changes for me. Still scratching my head on this one...

I'd just expect the Number of Fed Cells to reflect what I see in the log - if it is supposed to. It kinda seems like you would too?

[BrianL]

It does sound like a bug to me, you could try reporting it with IBM.

[Hippogriff]

On one hand I'm kinda hoping it's not just me... but, then, on the other hand I was hoping it was just my local environment that was awry and it was fixable as other folk would say "mine doesn't do what you're seeing".

[Steve Rowe]

The feeders that are in the rule sheet for Purchase, are they for the Purchase cube?

14980 [] INFO 2016-02-24 15:11:36.092 TM1.Server Feeders created in cube "purchase": new 3355, duplicate 0, ignored 0

could be

14980 [] INFO 2016-02-24 15:11:36.092 TM1.Server Feeders created "by processing the cube" "purchase": new 3355, duplicate 0, ignored 0

If the destination cells are not in the purchase cube then this could explain the different results in the front end?

[Hippogriff]

I've been advised that these two numbers should really match-up and that there are some fixes in this space in the latest release. I was asked to retry if I could do that.

[sumitgoyal2290]

I am facing exactly the same issue and wondering if you did get any solution or explanation for this scenario. "Number of fed cells" is shown 0 in cube }StatsByCube for all the cubes except for control cubes. It was working file before restating the server few days back.

[Bakkone]

Does your statscube show any value under "Memory used for feeders"?