How to Handle the log files that are generated due to index rebuild activity at source(ORACLE)

[Antony_05]

Hi Team,

We have a situation where the index is rebuild in source (ORACLE) in every weekend which causes the attunity replication with high latency. Since some Object ID are getting skipped in the transaction.

Appreciate any input if anyone has faced the similar issue in the past and found any solution. 

Thanks and Regards,

Antony S

[Steve_Nguyen]

If the maintenance activity performed during off hours then they can stop the task and start the task with timestamp after the activity(before business hours).

this is the best method.

[Heinvandenheuvel]

I wasn't aware of Index rebuilds slowing down Oracle replication significantly. I did know about this for SQLserver where the log reader may have to go back over the same log over and over every time having to adjust it object list in the select every time a new object is seen without a way to position.

How is the redo/archive log reading configured? You are not using to deprecated LOGminer are you? What Oracle version? Simple, or pluggable DB? Switch  logging to TRACE for PERFORMANCE (and leave it there for ever more) and evaluate what is being read/when.

Have you evaluated how much log (gigabytes) is being generated during the rebuild process?

Stop + start by timestamp allows one to completely skip the rebuild activity, but may cause dataloss.

Before you do that, why not try stop before index rebuild + resume after. That way the tasks can learn about all the new object ID's on resume, instead of one at a time. That helps for SQLserver - maybe also for Oracle.

hth,

Hein.

[Antony_05]

Hi @Steve_Nguyen , 

I will try the step that mentioned next week and update the result.

Thanks,

Antony S

[Antony_05]

Hi @Steve_Nguyen ,

In the above comment you have mentioned stop and start the task does this mean we have to stop the replication from Attunity end or in the DB end.

And also out of curiosity I want to know by stop and start of the task wont have any data loss.?

Regards,

Antony S

[lyka]

Hello,

the best method of rebuilding index would be:
a. Stop replicate task before rebuild index
b. Stop transaction log backup process before rebuild index
Or – stop MS-Replication process (see item 2 for more details)
c. Run rebuild index process
d. Once rebuild index done, let replicate consume all the changes.
e. Once replicate caught up with all the change and source latency is low,
Resume log backup (or MS-replication process)

Thanks

Lyka

[Antony_05]

Hi Lyka,

To be clear, Can you please let me know what is mean by MS-Replication process.

Regards,

Antony S