Had a wonderful Sangam16 conference in India, and received much applaud for the two presentations delivered, Oracle 12c multitenancy and Exadata migration best practices.
After a very short trip to India, life started to be business as usual again, and become busy. Was fully occupied with multiple assignments: Oracle EBS database health check assessment at a client for 2 days, GI/RDBMS/PSU deployments on Oracle Sun Super Cluster M7, Exadata configuration preparation and 9 databases migration to Exadata during the week-end.
Over the last week-end, we (me and my colleague) were pretty busy with 9 databases migration to Exadata. There were a few challenges , and learned a few new things too. I would like to discuss couple of scenarios that were interesting:
One of the databases had corrupted blocks, and the expdp was keep failing with ORA-01555: snapshot too old: rollback segment number with name "" too small. Our initial thoughts were tuning undo_retention, increasing the undo tablespace, setting an event, etc. Unfortunately, none of the workarounds helped in the situation. We then cameacross a MOS note which explains that an ORA-01555 with "", no rollback segment name is probably due to corrupted blocks. After applying the solution explained in the note, we managed to export/import the database successfully. My colleague has blogged about the scenario at his blog: http://bit.ly/2fBOxm7
Another database is running on Windows x86 64-bit, and its full of LOBs, hence, the datapumps (expdp) took significant time, as NFS filesystem used to store the dump file. We then thought of doing direct RMAN restore from source to target, as the database on Windows x86 64bit and Linux x86 64-bit are the same (Litten) Endian formats. As per one of the MOS notes, we can also do the Data Guard setup, and do RMAN restore. However, RMAN recovery would fail with ORA-600, as cross platform redo conversion won't be possible. We are now thinking of taking a cold backup (consistent) and do a complete restore with reset logs option.
Stayed tuned for more updates on this.
After a very short trip to India, life started to be business as usual again, and become busy. Was fully occupied with multiple assignments: Oracle EBS database health check assessment at a client for 2 days, GI/RDBMS/PSU deployments on Oracle Sun Super Cluster M7, Exadata configuration preparation and 9 databases migration to Exadata during the week-end.
Over the last week-end, we (me and my colleague) were pretty busy with 9 databases migration to Exadata. There were a few challenges , and learned a few new things too. I would like to discuss couple of scenarios that were interesting:
One of the databases had corrupted blocks, and the expdp was keep failing with ORA-01555: snapshot too old: rollback segment number with name "" too small. Our initial thoughts were tuning undo_retention, increasing the undo tablespace, setting an event, etc. Unfortunately, none of the workarounds helped in the situation. We then cameacross a MOS note which explains that an ORA-01555 with "", no rollback segment name is probably due to corrupted blocks. After applying the solution explained in the note, we managed to export/import the database successfully. My colleague has blogged about the scenario at his blog: http://bit.ly/2fBOxm7
Another database is running on Windows x86 64-bit, and its full of LOBs, hence, the datapumps (expdp) took significant time, as NFS filesystem used to store the dump file. We then thought of doing direct RMAN restore from source to target, as the database on Windows x86 64bit and Linux x86 64-bit are the same (Litten) Endian formats. As per one of the MOS notes, we can also do the Data Guard setup, and do RMAN restore. However, RMAN recovery would fail with ORA-600, as cross platform redo conversion won't be possible. We are now thinking of taking a cold backup (consistent) and do a complete restore with reset logs option.
Stayed tuned for more updates on this.
No comments:
Post a Comment