Rollbacks Observed While Executing INGEST

DB2 10.5 single partition HADR database on AIX 7.1 with a remote Runtime Client on Windows executing an INGEST for a 4 column, multi-row UPDATE. The input is a delimited file. There are no errors in the INGEST messages file and no rows rejected to the dumpfile. There are SQL0100W warnings in the INGEST messages file due to rows not found. The INGEST successfully completes and table data is as expected.

While using db2top to monitor I sometimes see the DB2_INGEST Application Status as Rollback Active or Rollback in progress. There are no deadlocks in the event monitor, no Lock timeouts, no increase in Lock waits and no Lock escalations in the database. Internal rollbacks are increasing but that also occurs while an INGEST is not executing. The db2diag.log shows STMM increasing Database_memory over the first few sequential INGEST executions but I have not observed STMM increasing Locklist. The INGEST rollbacks still persist when Database_memory reaches a steady state. The db2diag.log shows no errors while the INGEST is executing.

I tried tuning the INGEST with commit_count or commit_period and there appears to be no significant change in the rollback behavior.

Can anyone confirm the observed rollbacks are expected when using the INGEST utility?

If not, any comments on what may help reduce or eliminate these rollbacks?



Leave a Reply