Increase SQL Replication Frequency?

[Philip Bjorge]

Hi,

We're running the latest Raven 3.0 database and are interested in knowing how we can increase the frequency that documents are replicated via SQL replication?

On occasion, we're seeing delays of 3-10 minutes and it would be nice if we could have things consistently replicating every minute or so.

Are there any configuration settings we can tweak here?

Thanks

[Oren Eini (Ayende Rahien)]

This should be running constantly. There isn't any delay going on. 

It is possible that it is taking a long time because you are missing indexes?

Hibernating Rhinos Ltd  

Oren Eini l CEO l Mobile: + 972-52-548-6969

Office: +972-4-622-7811 l Fax: +972-153-4-622-7811

 

--
You received this message because you are subscribed to the Google Groups "RavenDB - 2nd generation document database" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ravendb+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Philip Bjorge]

We have indexes on the SQL table that is being replicated to and performance on that table isn't an issue.

Turning on debug logging shows that when replication is taking place, it happens very quickly.

I am seeing instant replication in our integration environment which has effectively no load.

Where SQL replication sometimes has 3-15 minute delays is in our production environment.

This environment has tens of millions of documents if that makes a difference.

It is a dedicated replication box and is not serving any reads or writes (other than through replication).

Looking at our monitoring, CPU load and memory consumption is low on the instance -- ~15% CPU consumption and ~50% memory consumption.

There are no replication failures.

Right now for example, the Raven/SqlReplication/Status document hasn't been touched in 10 minutes even though we've had hundreds of documents touched since then.

How would you recommend debugging this issue because I agree with you that looking at the bundle's code, this shouldn't be happening... but it is and isn't meeting our expectations of this feature.

To unsubscribe from this group and stop receiving emails from it, send an email to ravendb+u...@googlegroups.com.

[Philip Bjorge]

So our latest batch just went through and it replicated 10,000 items.

Below is a screenshot of the sql replication dashboard which shows a 10 minute gap between replication.

http://imgur.com/a/vaSgG

[Philip Bjorge]

So looking at the debug logs, there's just a legitimate 10 minute gap. I've "anonymized" the logs below.

2017-03-10 22:43:20.4442,Raven.Database.Bundles.SqlReplication.SqlReplicationTask,Debug,,78,"Replicated changes of ... for replication ...",

           # We have about 10 other things being replicated here 

2017-03-10 22:44:07.2555,Raven.Database.Bundles.SqlReplication.SqlReplicationTask,Debug,,119,recorded a deleted document ...

# Lots of deletes recorded here

2017-03-10 22:53:29.4755,Raven.Database.Bundles.SqlReplication.SqlReplicationTask,Debug,,240,"Replicated changes of ... for replication ..."

           # We have about 10 other things being replicated here 

 

Thanks for looking

We are running build #30171

[Oren Eini (Ayende Rahien)]

Can you send us the full log file (please include the customer details as well in the email to sup...@ravendb.net).

To unsubscribe from this group and stop receiving emails from it, send an email to ravendb+unsubscribe@googlegroups.com.

[Philip Bjorge]

While gathering our logs, we determined the issue.

For one of our collections, a document change could trigger 10s of thousands of inserts... Since these aren't happening as bulk inserts, even though they each took only 1ms, they seem to stop up replication.

We've resolved by disabling this replication collection for now, with a more permanent fix being to split up into two sql replication instances.

Has the behavior in 3.5 changed at all such that this might not be an issue?

Thanks

[Oren Eini (Ayende Rahien)]

No, this didn't change in 3.5

What scenario do you have that a single document will have so many inserts?

To unsubscribe from this group and stop receiving emails from it, send an email to ravendb+unsubscribe@googlegroups.com.

[Philip Bjorge]

A pretty degenerate scenario to be fair.

Large documents (1+ MB) that had large arrays of child objects in them. These child objects were being replicated to their own table.