Cubes get empty after Windows restart. How to fix this?

[abcuser]

Hi,
I am using Cognos TM1 v9.5.2 on Windows 2008 RC2. When I restart Windows I have noticed that some of the cubes on TM1 server loses all of the data (cubes are empty when I query them) and some of cubes do not lose data (which is OK). According to documentation when TM1 Windows service is stopped all of the data should be saved to disk. This is not a big problem, because I can fully reload all cubes from warehouse, but it is annoying and time consuming. I am just wondering why does this happen on some cubes and it does not happen on all of them? When ever I restart Windows I would like that ALL of the cubes preserves data. Most probably there is some setting that is specific to cube. Is there a settings specific to cube? Which one?

P.S. This problem was already there on some cubes before we migrated from v9.5.0 to v9.5.2. So most probably some setting from "old times".
Thanks

[lotsaram]

This would be because ....
1. When you load from the DWH transaction logging is switched off (sensible)
2. There has been no SaveDataAll executed after loading the cubes before the server is brought down

The solution is simple just ensure that a data save is scheduled or run manually BEFORE a reboot of the physical server. If the machine is simply restarted and the TM1 server has a lot of data to save then windows will simply conclude that the service is hanging and terminate it before all cube updates can be saved to disk which would be my guess as to what has happened in your case.

[abcuser]

Thanks a lot of answer.

1. When you load from the DWH transaction logging is switched off (sensible)

What is DWH? How can I check transaction logging? Is this logging turned on for particular cube? When ever I restart Windows I always get two cubes that are empty and all of others are loaded with data.

2. There has been no SaveDataAll executed after loading the cubes before the server is brought down
The solution is simple just ensure that a data save is scheduled or run manually BEFORE a reboot of the physical server.

If I understand correctly I just create a new process in TM1 Architect inside Turbo Integrator with command "SaveDataAll" in Prolog tab. Is it?

[Christopher Kernahan]

DWH = DataWareHouse.

Transaction logging is a setting for each cube that will log every change made to values in that cube by user and timestamp.

From the documentation, "the Logging property is set in the TM1 Security Assignments dialog box and stored in the }CubeProperties control cube." See also CubeGet/SetLogChanges.

Typically when loading very large amounts of data it is turned off in the TI for the duration of the dataload and then turned back on by the TI when this is finished.

This is to prevent cluttering up the Transaction Log when you know that all those transactions would have been executed by TI, allowing it to be more useful to you in circumstances when you are searching for a single user's changes, or changes to particular parts of cubes.

With regard to SaveData, we typically setup a daily chore to run the SaveDataAll command in the Prolog as you say. Be careful to ensure this won't conflict with other Chores/processes.

[Alan Kirk]

Someone got a post through ahead of me covering some of these points, but since I've typed it out I'll post it anyway...

Thanks a lot of answer.

1. When you load from the DWH transaction logging is switched off (sensible)

What is DWH?

Data Warehouse.

How can I check transaction logging? Is this logging turned on for particular cube?

Right click on the Cubes icon in Server Explorer and select Security Assignments.... That will show the logging status for each cube.

However when you're loading from a ledger or DWH system it's more typical to have the TI process temporarily turn off logging via the CubeSetLogChanges function. You use that to turn logging off, zero out the data, reload the data, then use the function to turn logging back on. However as Lotsaram was was explaining, if you go down that path you do need to (or more precisely, should) do a data save regularly, long before you restart the server, much less Windows itself.

Do not rely on either shutting down the service, or worse, shutting down Windows, to do that save for you. Yes, in theory it should. But in practice if you have a slow hard disk system or really large cubes the service may well spit the dummy and claim that TM1 is not responding "in a timely manner" and give up on the save before it completes. ALWAYS force a save using one of the methods below BEFORE you do a shut down.

When ever I restart Windows I always get two cubes that are empty and all of others are loaded with data.

2. There has been no SaveDataAll executed after loading the cubes before the server is brought down
The solution is simple just ensure that a data save is scheduled or run manually BEFORE a reboot of the physical server.

If I understand correctly I just create a new process in TM1 Architect inside Turbo Integrator with command "SaveDataAll" in Prolog tab. Is it?

You could do that, or if you want to do it manually just right click on the server icon and select the Save Data command.