Cannot Drop or Rebuild an Index - set index flag error

[Former Member]

Hello Experts. I am running into an error while dropping or rebuilding an index. any help on how to resolve this or how i can get rid of this index is really helpful. this is an index created on a column store table with 16.5B records for some performance reasons. Now i get random execution plan error on the procedures written on top of this table(have an analytic view on this table) for data extraction. this is so random and i could find below in the trace files along when the execution plan error was throwed. so not sure if this has to do any with this corrupted index.

BwPopAggregateParallel.cpp(01053) : Index syncpoint mismatch;object=

Below is the error i get when i try to rebuild the index.

ALTER INDEX IDX_XXXXXX REBUILD;

Could not execute 'ALTER INDEX IDX_XXXXX' in 285 ms 613 µs .

SAP DBTech JDBC: [2048]: column store error: set index flag error:  [2999] general error (no further information available)

HANA is on revision 73

Any help / pointers?

-Thanks,

Vinod Nair.

[lbreddemann]

1. This should be dealt with in an support incident.

2. Why are you rebuilding a column store index?

[Former Member]

Hi Lars, Yes we have created an OSS message with SAP to troubleshoot this. on a seperate note the execution plan error is happening when a call to procedure happens in parallel on 200+ threads. so far we were in a sweetspot on 120 threads running in parallel for select where we didnt get this error. not sure what is adding overhead on increasing the threads for select. My was guess was a table search error and somehow this corrupted index adding to the error, not 100% sure.

I am trying to get rid of this index so even a drop index gets me to the same error

Could not execute 'DROP INDEX IDX_XXXX' in 1:49.014 minutes .

SAP DBTech JDBC: [2048]: column store error: set index flag error:  [2999] general error (no further information available)

[lbreddemann]

Sounds like any exclusive index access (like the one required to drop or rebuild it) seems to fail.

Did you manage to unload/load the table?

Concerning the amount of threads: I am not aware of any dependency between "index present" and "number of threads".

Really, we miss information here and the support incident is - in my eyes - the best way to find all required information.

Would be nice if you post the result of that!

- Lars

[Former Member]

Hi Lars, just wanted to update on this issue. here is how the index error fixed. the table was partitioned to 16 nodes and one of the blade went down on a fail over scenario. after redistribution the index was corrupted either cant rebuild it nor delete it. this was showing the error i posted initially. Once the blade was up and when we re distribute the tables,i was able to drop the index with out error.

Also on the execution plan error, we were able to put 240 threads in parallel for select and with help of product support team we identified it as an error in the executor.ini time out configuration. we were able to get rid of the error with some different time out settings in executor.ini.

-Thanks

VInod Nair