How to react on the “bia_indexes_size_perf_check” alert in BWA Monitor
How to react on the “bia_indexes_size_perf_check” alert in BWA Monitor
While checking in BWA Monitor, RSDDBIAMON2 we may come across the alert “bia_indexes_size_perf_check: Returns a list of BWA indexes”. This occurs when physical index has reached the threshold and this scenario usually develops when the index is created on a single BWA blade instead of the all available blades.
To find out the cubes for which index have reached the threshold value click on the details tab. Here we can see the Indexes which are not split. In this case the cubes are Y*XP2**0 and Y*XC5**0.
We can also get details on the type of index and the recommended number of partitions that can be used to split the index. In our case we are having 28 blades available and so in the details it shows recommended number of partitions as 28. Also note that the index type is “Fact Table”
We can confirm the above information by crosschecking in TREXADMIN. Provide the name of the system you want to connect to.
In TREXADMIN go to Index Landscape tab and provide the Fact table names collected from RSDDBIAMON2 by adding “*” at the beginning and end and click display icon. In our case fY*XP2**0 and fY*XC5**0 with * added at beginning and ending.
Here we can see that the index is created only on a single blade eventhough we have another 27 blades available for splitting and distributing the index so that the threshold value will not be exceeded.
So now we have to recreate the index by distributing it on all the available blades. To do this, go to RSDDV and give the cube name and go to BWA Index maintenance.
Click on Delete to delete the existing index and when it pop up for confirmation click Yes. After this again we will receive a pop up asking for confirmation to delete the Master Data Indexes. As Master Data Indexes are shared between cubes referring to same Master Data it is advisable not to delete this.
Now select “Create” option to recreate the deleted index.
Go to the “Load Admin” tab and by default the selected setting will be “SYSTEMEINSTELLUNGEN” which means SYSTEM SETTINGS. Change this and select the option “EIGENE EINSTELLUNGEN” which means My Settings. Also set the number of records to some high value say 99999999.
Now select “Activate and Fill BWA Index and select immediate option. This will schedule a job with name “RSDDB_BIAINDEX_FILL”
Monitor the job “RSDDB_BIAINDEX_FILL” to make sure the reindexing is successful.
Once this indexing job is completed again check the “Index Landscape” in TREXADMIN and we can see that the indexes are splitted on all available blades.
On Refreshing we can see that the alert does not exist anymore. Splitting the indexes on different blades helps to optimize the performance.
Hi George,
Are you on BW7.0x or BW7.3?
Hi Ali,
We are on BW 7.3.
Hi George,
at first thanks for your blog. This is one of the most frequently asked questions regarding BWA.
With your solution you have to reindex the complete index. In some cases this will take a lot of time and is not possible during normal operating time.
For this you can split it with the python standalone tool:
→ tab Landscape →search for the index → right click on Split/Merge Index → “ROUNDROBIN x″ x is the number of active blades (includes not the backup blades) it splits the selected index in x parts. The number of parts should be always the number of active blades to ensure the best performance for searching for this information. During the time index is read-only!
Regards,
Jens
Hi Jens,
Thanks for the information. As we are not using Python in our environment I was not aware of this.
Regards,
George
Hi Jens,
We have also followed that kind of reorg in order to solve this problem but the alerts are not turned off. Do you know how can we turn the alterts off after the cube has been correctly reorgnized with python tools?
Regards,
Eumir
Hi George, thanks for the blog post.
From my experience, BIA indexes need to be dropped and rebuilt on a regular basis (weekly) to optimize the performance of the blade servers.
Hi HS Kok,
You are correct. The best practice is used to drop the indexes and rebuild it weekly. But due to any activities if the indexes of a particular cube is rebuilt and if we face the issue we can proceed with this practice. Else it is always recommended to do the index maintenance weekly by scheduling a PC.
Regards,
George
HS,
Your comment about rebuilding once a week - I assume that's for delta indexing? We drop our indexes as part of our daily process chains - we don't use delta indexing. Just checking if I should build a new PC 🙂
Thanks
Hi Cobus,
Not sure what delta indexing you are talking about. I'm referring to the BIA indexes, not the InfoCube index.
When you load delta requests to an InfoCube, a BIA index roll-up needs to happen before you can see the delta information in your BW reports. In such cases, the BIA indexes should be fully dropped and re-created on a weekly basis as the performance will degenerate over time.
Hi,
We have the same warning for X tables and S tables, how do we rebuild the Master Data tables? We are on BW 7.0.
Hi SAm,
You can drop the index for the entire cube using RSDDV and then rebuild using the required settings, split to all the available blades.
Have you tried program: RSDDTREX_SINGLE_TABLE_INDEX
We have used this in the past and it's worked perfectly for what you're describing.
-Rob
Founder of SAPBWTraining.net
Handy information,keep it up.
Hi all,
Please also check SAP note "1791048 Bia_indexes_size_perf_check Alert" for more details about this issue.
-Diego Seben