Please enable JavaScript to view this site.

2020/12/07

•Db2/z Log Capture (SQDDB2C): Capture will be more responsive when in a caught-up situation. When the Db2/z Log Miner returns a complete buffer more data is immediately requested. When there are no outstanding UOW's to publish and no new data returned by the Db2/z Log Miner, the capture gradually extends the wait period between Log Miner requests. While that will still occur, it will be more aggressive and use a shorter initial wait.  Previously if a partial buffer was received, Capture would wait 3 seconds before requesting more data from the log, and if no data was received, increase the wait request delay by 1 second with a 5 second maximum wait. The initial wait following a partial buffer was reduced 250 ms, followed by 1 second increments up to the maximum 5 second wait. The change provides more sub-second latency on lightly loaded systems

•Oracle LogMiner Capture (SQDLOGM): Corrected corner case in a very low activity RAC environment where the Oracle SCN (system change number) of the last log record processed has still not advanced by the time the Capture makes a subsequent request to the Oracle LogMiner.

•Replicator Engine (SQDRPL): Corrected key-hash inconsistencies in JSON/AVRO topics that resulted in different partition assignments for insert/update/delete activity for records having the same source key(s).