BW Accelerator Upgrade: P2: Practical Usage Scenario of back up modes
In this blog, I would like to share my experience during BWAccelerator Upgrade which helped me in understanding practical usage, pros/cons for different BWA back up modes.
I have also shared one of my experiences on BW Accelerator Upgrade in BW Accelerator Upgrade: P1.If you haven’t already read this blog please consider this for reading.
After the BWAUpgrade, when we went ahead to rebuild the BWA indexes of the cubes, theindexes were dropped properly but while creation/filling, they started failing.
The error messageswe got were:
“Creation of Logical Indexes ‘ZXXC01’terminated. Error text join index create error; index=<sid>_zxxc01,location=biablade7:31003 2433″ and<br />”Creation of BIA index for Info Cube ‘ZXXC01’ terminated during activation”
Since the errors were pointing towards BWA blade 7, we analyzed the traces for this. Onanalyzing in detail , below messages were retrieved.
“DeleteIndex: Locking for index Id:<sid>_zxxc01, failed.”
“TRexApiJoinIndexCreate.cpp (00250) : errorin creating join index”
We restarted the services of blade 7, but still no luck. Now we contacted SAP, SAP suggestedthat there seems to be some communication issue that could be or/not the causeof indexes error. As per SAP, most likely some indexes have not been successfully deleted and then there is a lock problem when trying to create another one. But to get a clearer picture, SAP suggested doing a physical restart of all the BWA blades.
Though from our analysis we were almost sure that there is no communication issue, but since it was our production environment which was impacted and both the parties were not sure about the root cause of this issue, we restarted all the blades, but again got the same set of errors. Then SAP suggested doing a reorg and then creating the indexes.The idea was, this reorg will balance the indexes between the blades and willalso remove inconsistencies if any. Reorg was done (TCODE TREXADMIN ) Advanced Mode), but this time as well luck was not with us.
Then it wasinvestigated and found that indexes werefailing during the phase “Initial Reorg” and the backup mode we were using was mutual. Now, the interesting fact was, there is a known issue for Initial Reorg with backup mode as mutual.SAP also confirmed, that after Revision 49 there were some customer messages with similar issues with join indexes which were resolved by changing the backup mode.
Then, we took the help of SAP Note 1383228 to decide if not mutual, and then which back up modeis recommended for our latest revision.
Some good pointsobserved were:
1. It isnot recommended to use backup modes mutual in productive BW systems.
2. Forsmaller BWA landscapes, use of backup mode shared is recommended.
3. Forlarger BWA landscapes, use of backup modes multiplexed with several backupblades is recommended.
Our being a larger BWA landscape, finally it was decided to use multiplexed back up mode. And this really helped as it solved the problem and it was a good learning for all my project mates.