hana - may significantly degrade performance

[Former Member]

I found traces below in index server trace . anyone hit similar issue below . This only occur  following a cluster fail over for hardware replacement

Potential performance problem: Table SAPABP:/BIC/AZFMIAD0700 is split and table SAPABP:/BIC/AZFMIAD0740 is split by an appropriate criterion but corresponding parts are located on different servers. This will not prevent the Data Store from working but it may significantly degrade performance.

[60708]{-1}[-1/-1] 2014-10-13 09:53:50.832197 e Statement

mvcc_anti_ager.cc(01136) : There are too many un-collected versions. ('number of versions for one of the partitions > 1000000' or 'maximum number of versions per record > 100000')

The transaction blocks the garbage collection of HANA database.:

CONNECTION_ID = -1, HOST = kulaphana01:30003, VOLUME_ID = 5,

TRANSACTION_ID = "315", TRANSACTION_TYPE = "EXTERNAL TRANSACTION",

PRIMARY_TRANSACTION_ID = "47", PRIMARY_TRANSACTION_TYPE = "USER TRANSACTION",

UPDATE_TRANSACTION_ID = "0", MIN_MVCC_SNAPSHOT_TIMESTAMP = 762429068,

TRANSACTION TOTAL EXECUTED TIME = 146845 sec

[michael_gerstmayer]

Hello,

Apparently there is an inconsistency in the partitioning of the BW table which should be repaired. Can you please implement the latest version of report RSDU_TABLE_CONSISTENCY (see SAP Note 2025271) and run it with option 'Check the Partition Specification of tables'? It will detect the inconsistency in the partitioning definition which can be repaired by report.

The MVCC issue is a transactional problem where the garbage collection of a row-store table is blocked (e.g. because another transaction did not yet send the "commit" on updating the table, thus older versions need to be kept). To identify the "blocker" and resolve the situation, please follow the SAP HANA Troubleshooting and Performance analysis guide, see the corresponding chapter:

Multiversion Concurrency Control (MVCC) Issues - SAP HANA Troubleshooting and Performance Analysis G...

Due to the failover (of the application servers?), it might be that the transactions aborted prematurely before sending the commit. Idle/Dead connections possibly blocking the garbage collection will be terminated by HANA itself (when the idle_connection_timeout or tcp_keepalive* is reached, see SAP Note 1836774 for details).

Best regards,

Michael

[former_member183326]

Could you please try to restart the Xsengine. If you are unsure how to do this please refer to SAP KBA 2087751