What is Latching in Qlik replicate?

[suvbin]

what is latching in the Qlik replicate.  At the sql server db side, they are facing high OS paging issues. What should be done at this time at the Qlik replicate side.

[Heinvandenheuvel · Accepted solution]

I have no clue what you mean with 'latching'. The word has not meaning in Replicate as such best I know.

The only somewhat latch related explicit operation Replicate uses is managing the 'attrep_truncation_safeguard' table if that is even selected impacting 'repl_done' and active log truncations.

There may well be implicit latch activity from statements Replicate executes on the source, but you'll have to tell us what you believe to see which connects back to Replicate.

Please use more words, copy and past some examples from what you believed shows a problem.

>> At the sql server db side, they are facing high OS paging issues

Typically that is overcommitted memory.

Are we supposed to guess here that this started happening when certain Replicate activity was started.

There are numerous tools built-in and 3rd party both under windows (Process Explorer, Perfmon,...) itself and session memory usage tools in SQLserver (SSSMS) to highlight what might contribute to paging and/or excessive memory usage.  If it has to to with Replicate, then you may want to isolate the statements involved [select (dump)_db_log?).

>> What should be done at this time at the Qlik replicate side.

How can anyone help with that if you do not bother to attempt to describe what Replicate is asked to do. Perhaps a (sanitized) task export can help some, but you words to describe what is trying to be accomplished and how that is believed to cause issue would be better.

hth,

Hein.

[lyka · Accepted solution]

Hello,

With regards to the table mentioned by Hein,

In SQL Server, We keep open 2 transaction as part of Replication. Please refer to the source table : attrep_truncation_safeguard

There will be 2 rows per active task: A and B
Replicate opens a RW Transaction and selects one of the row, after 5 minutes (Endpoint config), it starts a TX for the other RX, updates the first row to the current time and commits. Thus there is always an open transaction.

You should not run any maintenance against the attrep_truncation_safeguard table. If you cannot exclude this table from the maintenance plan, then you should stop the Replicate task during the maintenance window.

Hope this helps!

Thanks
Lyka

[suvbin]

Where can i find this " attrep_truncation_safeguard" ? in the source db?

[Heinvandenheuvel]

>> Where can i find this " attrep_truncation_safeguard" ? in the source db?

DONT. Not now, not yet.

First Tell us what YOU KNOW. What you see, why you are asked the main questions in the first place.

Right now, for me, your main question is similar to 'my car makes a funny sound, how do I change the trailer?' Followed by 'where is the caliper?'.

Try to restate you question. Help us help you.

Hein.

[lyka]

Hello.

Its in the source database  - if you are referring to the latch in SQL server.

Thanks

Lyka