Server crash when users write during TI sequence run
Posted: Thu Jun 03, 2010 6:01 am
Hi,
We have experienced multiple crashes of our production server today and have tracked it down to the following sequence of events, but are at a loss as to how to prevent it.
Note: We are running TM1 9.1.4 x64
A user needs to manually run a Turbo Integrator process at any point during the day that it set up to run a sequence of sub-TIs via EXECUTEPROCESS functions.
These sub-TIs run against varying cubes as source data, and unfortunately cannot be combined into one TI.
If another user writes values to any cube at the same time as the process is initiated, there is a write-lock when a process is running, but during the gaps between the sub-TIs being executed, the write-lock is released and the server crashes.
On the advice of IBM we have tried the following - all to no avail.
1) Turned logging on ALL cubes off at the start of main TI (and back on at the end)
2) Turned on Batch updating via a BATCHUPDATESTART and BATCHUPDATEFINISH in the main TI
3) getting the user to manually run a chore instead of a the TI directly.
We are 99% sure that in TM1 version 9.4, putting EnableBulkLoadMode() at the beginning of the main TI and DisableBulkLoadMode() in the Epilog will lock all users from all cubes for the duration of the entire TI process sequence and resolve the issue, but the upgrade is a few months away and we need a solution for TM1 9.l.4 now.
Do any of the giant brains here have any ideas on what we can do to prevent this, or have ideas on different methods to avoid the issue all together?
Many thanks
We have experienced multiple crashes of our production server today and have tracked it down to the following sequence of events, but are at a loss as to how to prevent it.
Note: We are running TM1 9.1.4 x64
A user needs to manually run a Turbo Integrator process at any point during the day that it set up to run a sequence of sub-TIs via EXECUTEPROCESS functions.
These sub-TIs run against varying cubes as source data, and unfortunately cannot be combined into one TI.
If another user writes values to any cube at the same time as the process is initiated, there is a write-lock when a process is running, but during the gaps between the sub-TIs being executed, the write-lock is released and the server crashes.
On the advice of IBM we have tried the following - all to no avail.
1) Turned logging on ALL cubes off at the start of main TI (and back on at the end)
2) Turned on Batch updating via a BATCHUPDATESTART and BATCHUPDATEFINISH in the main TI
3) getting the user to manually run a chore instead of a the TI directly.
We are 99% sure that in TM1 version 9.4, putting EnableBulkLoadMode() at the beginning of the main TI and DisableBulkLoadMode() in the Epilog will lock all users from all cubes for the duration of the entire TI process sequence and resolve the issue, but the upgrade is a few months away and we need a solution for TM1 9.l.4 now.
Do any of the giant brains here have any ideas on what we can do to prevent this, or have ideas on different methods to avoid the issue all together?
Many thanks
