TM1 Forum

Hello everyone,

I am experiencing an "issue" with SaveDataAll and I was not able to find anything in both help and this forum.

I have assigned subsets to views in our model and ran SaveDataAll process manually (it is also scheduled to run a few times a day) to make sure that metadata changes are saved immediatelly.
After a few days our server crashed for some reason and after restart the subsets were not assigned to views.

I have also tested in my local environment (admin server is running as an application in this case). I have assigned subsets to views => ran SaveDataAll process manually => shut admin server down without saving data (to test SaveDataAll) => started admin server => subsets were not assigned

I suppose SaveDataAll should be enough to save metadata changes... Am I missing something (settings in configuration file,...)?

Thank you for help.

Stanislav

I have assigned subsets to views in our model and ran SaveDataAll process manually (it is also scheduled to run a few times a day) to make sure that metadata changes are saved immediatelly.
After a few days our server crashed for some reason and after restart the subsets were not assigned to views.

Hi Stanislav,

SaveDataAll is meant to save data (cell-value) changes. Metadata changes are saved instantaniously.
If in your case the metadata changes are not saved directly - and your server crashes for 'some' reason - perhaps there's something wrong with the permissions of the user the TM1 server is running on or maybe the (some) files of the TM1 database are set as Read-Only.

Michel

SaveDataAll is meant to save data (cell-value) changes. Metadata changes are saved instantaniously.

Hi Michel,

Thank you for your answer. I am just a bit confused now.
Below is what you can see in TM1 help. Does it mean that when I assing subset manually it is saved immediately and if done through TI then it is not?
Help says you can use SaveDataAll in the Epilog to save metadata changes. Does it have to be included in the same process which is updating Metadata?
""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
SaveDataAll
This is a TM1 TurboIntegrator function, valid only in TurboIntegrator processes.

This function saves all TM1 data from server memory to disk and restarts the log file.

Important: Metadata changes made in a TurboIntegrator process, such as assigning a subset to a view or adding an element to a dimension, are not automatically saved when the TurboIntegrator process completes. You must explicitly save your changes or the changes may be lost when the TM1 server is shut down.

To save metadata changes, you can use the SaveDataAll function in the Epilog procedure of the TurboIntegrator process.
""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""

Stanislav

Hi Stanislav,

It appears that the 'important' statement regarding metadata changes is listed in the 9.4MR1 documentation, but not in the 9.5.x documentation...

After running some tests (on 9.5.1) I conclude that the metadata changes I tested were saved immediately, except for view/subset assignments. These were saved to disk after running the SaveDataAll. I wasn't aware about this.

Have you had a look at the server log to see whether there are error messages regarding SaveDataAll - I know there are or have been problems with this function. F.i. locking problems can lead to chore roll-back situations or even server crashes - maybe such a situation was also causing your server crash?

Michel

Michel Zijlema wrote:Hi Stanislav,

It appears that the 'important' statement regarding metadata changes is listed in the 9.4MR1 documentation, but not in the 9.5.x documentation...

After running some tests (on 9.5.1) I conclude that the metadata changes I tested were saved immediately, except for view/subset assignments. These were saved to disk after running the SaveDataAll. I wasn't aware about this...

This is a very interesting titbit of information, it is certainly something I also wasn't aware of. As a rule all meta data changes in TM1 are committed to disk immediately. Possibly a workaround in this case would be to destroy and recreate the views rather than just update them you might then get immediate commitment of the new view definitions (would have to test and see) ...

It does seem absurd that a save data all is needed in order to force writing up to date subset assignments to views as a data save in TM1 is otherwise exclusively for cube cell data.

Michel:
Have you had a look at the server log to see whether there are error messages regarding SaveDataAll

Hi Michel,
I have checked the log - there are no messages related to SaveDataAll in this case. I have not seen any at all yet (for approx 1 year).

Lotsaram:
Possibly a workaround in this case would be to destroy and recreate the views rather than just update them

Hello Lotsaram,
Thank you for help. I will surely give it a try if there is not other solution.

Michel:
perhaps there's something wrong with the permissions of the user the TM1 server is running on or maybe the (some) files of the TM1 database are set as Read-Only

As written above, there are no errors in the log. But while process is running (approx 1or 2 seconds) Process Progres Reports similar to the one below are displayed:
"Completing Could not write dimension definition to disk storage. procedure of process POR input GP"

where "POR input GP" is a subset within dimension. There are more messages like this one for other dimensions, too.

It seems user permissions might be the case. I am going to investigate with our server administrator and come back as soon as resolved.

Thank you for now.
Stanislav

sbalaz wrote: It seems user permissions might be the case. I am going to investigate with our server administrator and come back as soon as resolved.
Stanislav

Service account belongs to local administrators group and the views files are not read-only - this looks OK.
SaveDataAll in the epilog of the "subset assigning process" did not work as described in documentation.

Stanislav