TM1 Server crash on saving dimension (status Suspended)

[Ken Vuong]

Hi everyone,

It's been a while since since I posted anything. Hope everyone is doing splendidly well!

We have recently been getting intermittent PA server (11.0.5.79-0 on cmplst file) crashes when saving dimensions in Edit Dimension Structure.
Our dimension are rather small but when it happens, the status on the PA server goes into 'Suspended', the ram would drop then increase like it was trying to reload and then tm1sd.exe disappears.

There is nothing in the Tm1server file, the model is tiny and there is ample of Ram (server loads at 1GB. 5GB used. 58GB available)
I have moved the model to Dev and it works it annoyingly work fine every time.

Has anyone encountered anything like this?

Many thanks in advance,
Ken

[lotsaram]

This is common when using the Architect dimension editor with a v11 server. Especially if the dimension has alternate hierarchy.

[Steve Rowe]

Hi Ken,

We have seen this a few times as well.

The only repeatable instance we have had is when two people were in the perspectives dimension editor at the same time and had both performed edits (obviously not best practice but it does happen). We have raised this with IBM. (TS001876885)

We have instances of server crashes when using the perspectives editor without two people being in as well though, nothing we have been able to reproduce so nothing with IBM as yet.

So you are not alone...

Cheers,

[Ken Vuong]

Thanks guys.

Good to know it's not only me. Looks like I will try to do this after hours or setup a TI to do it.

Our reseller (sempre) suggest playing around with some CFG parameter to see if that helps.
I will update if I can re-produce this problem and get something concrete to IBM.

...feels like we going back in time to the TM1 Version 6 days! lol

Ken

[PavoGa]

Hi Ken,

We had the same thing and were able to isolate at least one cause which was 100% repeatable.

When loading data, if there is an element that does not exist in a dimension, we add that element to a consolidation called "Unregistered" to deal with later. If that consolidation is the first element in the dimension, when we open the cube, it crashes the server (may have to close it once and then reopen it, but never more than twice opening will cause the crash). The dimension with this apparently suicidal structure has to be the first dimension in the cube.

Sounds nuts, but can do it right now and cause a crash. Do not recall if we tried another name for the consolidation because once we confirmed this, simply made sure "Unregistered" showed up at the bottom and problems went away.

[Steve Rowe]

Hi PavoGa,

What's your version number please? Did you get this acknowledged by IBM?

Cheers,

[PavoGa]

Hi PavoGa,

What's your version number please? Did you get this acknowledged by IBM?

Cheers,

Yes, we did. PA 2.0.05

[Steve Rowe]

In response to our defect we've been told that there was a defect relating to "re-ordering elements" in the dim editor (rather than formally setting the sort order) causing the server to fail.

This was fixed in 205 IF 3 and 206

We tested our use case and have confirmed the issue is resolved.

Since nearly any change to a dimension must involve reordering the list in some way I'm guessing that this may solve many peoples issue.

APAR for the resolved defect was PH01140
HTH

[PavoGa]

Thank you, Steve. When I get a chance, I'll try it again.

[Ken Vuong]

Thanks Steve.

I will upgrade to 206.
Cheers,

Ken

[Wim Gielis]

Thanks Steve.

I will upgrade to 206.
Cheers,

Ken

Anyone has an idea when 2.0.7 will be out ?

[Ken Vuong]

Just an update on this log.

Finally updated the server to PA 207 to go to PAW, and the bug has gone. Server no longer crashes on saving Dimension.

Would recommend anyone who is on PA 205 to move away from it if possible.
Cheers,

Ken

[Steve Rowe]

Wow, that took a while to get organised, Merry Christmas Ken!

[paulsimon]

Hi

It looks like Ken's problem was specific to the Dimension Editor, but from PavoGa's description it looks like a crash from a TI process. Is that the case?

We have also had some crashes with the Dimension Editor but fortunately only in Non-Production environments, as we don't really use the Dimension Editor in Prod.

We are on 2.0.5 and just had a crash in Production yesterday. There was nothing in the TM1Server.log at all to indicate the cause. There seemed to have been no activity at the time. It was late in the day and the only activity was a user logging chore, which according to the TM1Server.log finished executing successfully a few seconds before the crash.

I couldn't tell you what fix pack we are on. The version number on tm1sd.exe says 11.3.0.27, but I have long since given up trying to make any sense of IBM's approach to version numbering.

For the moment, we have reported the crash to IBM and put in additional logging to capture a crash dump if it happens again. Hopefully, it won't as 2.0.5 has actually been reasonably stable for us. We do restart the service every night, which may help.

We would be reluctant to upgrade without a full system test because of the risk that 2.0.7 cures one bug and introduces three brand new ones. At present we have a lot of projects on the go and don't have the capacity to carry out a full system test. In any case, we are going into Interims followed by Year End so there is no chance of us getting permission from our users to upgrade, or any resource to carry out the user side of testing.

So really just interested to hear if anyone has had crashes on 2.0.5 that were not attributable to the Dimension Editor?

Regards

Paul SImon

[Ken Vuong]

Thanks Steve.

Wishing Happy Christmas to you and the Infocat guys. Have a splendid New Year too

Cheers,

Ken