TM1 Forum

Hello,

In one of our large PROD TM1 servers, the TM1 server keep crashing intermittently when a user tries to open views, especially when opening larger cube views that exceed view memory limit. Crash happens almost every day. Anyone else observing similar issues?

Since the server is too large, I cannot send its dump to IBM Support and we cannot even get permission for this.

I have the following config in this TM1 server?
MTFeeders=T
MTCubeLoad=T
MTFeedersAtStartup=T
MTQ=-1

Any comments would be appreciated.
Regards,

Are there any messages in the log or are just just maxxing out the RAM?

There was a regression defect in 2.0.9 fixed in 2.0.91 relating to stargates that was crashing the box that one of our customers hit but AFAIK it was not present in 2.0.7, this generated a DEBUG message in the log even when no extended logging was turned on, so if you are hitting this there should be a smoking gun.

If you need to minimise RAM consumption you'll need to turn the the three multi-thread at start-up options off and live with the start-up time, perhaps adding persistent feeders to claw some time back.

If it's not just RAM load and you have no messages to give us a clue, maybe turn some of the extended logging on to get a hint. Perhaps the stargate ones?

There is no special messages in the tm1server log. There is no message about memory issue. No errors at all. It suddenly goes down. In TM1Top, we can see that the last operation was a cube view being oped by a user. Issue is not specific to any cubes or any views. It intermittently happens when a user tries to open a large cube view that exceeds view memory limit. Since this is a large tm1 server (>500GB memory), I cannot reduce MTQ parameters I think, and I cannot enable debug logging since this is a very busy tm1 server. What specific logging options that you can suggest enabling?

Hello!

1. Try to increase maximumviewsize parameter
2. (if Windows) Does Event Viewer say something useful?
3. Is it possible to upgrade to 2.0.9.1? Or at least to .08, considering last news regarding Security and dynamic subsets... They have 5-10 issues with "server crash" fixed every fixpack. Maybe one of them is yours
4. Can you reproduce this? You can create a new server (with for example 1/10 of cubes), obfuscate data, reproduce the crash and send the dump to IBM