RavenDB 3.5 replication configuration

47 views
Skip to first unread message

Nick

unread,
Jun 22, 2016, 6:32:13 AM6/22/16
to RavenDB - 2nd generation document database
Hi,

I have 3 questions on the above...

1. Client Failover Behaviour 

I have a 3 node cluster on 3.5, my understanding is that there will always be a nominated leader in the cluster which accepts all writes, other nodes will accept read requests, should the current leader fail another leader will be selected and will remain the leader, clients will now send all writes to the new leader.

That being the case I am confused about what we should specify for the Client Failover Behaviour in the global replication configuration, I am presuming we should leave this as "Let Client Decide" and the client will behave in the correct way. Is that correct?


2. Document store setup

Is it still recommended in 3.5 to specify the failover servers when creating the document store in our clients, assuming we're running a cluster?


3. Transitive Replication Behavior

I am also not clear on what we should be specifying for the TransitiveReplicationBehavior on the replication destinations.

Thanks in advance

Nick


Tal Weiss

unread,
Jun 22, 2016, 7:01:39 AM6/22/16
to RavenDB - 2nd generation document database
1. Yes i think leaving it on the "Let Client Decide" is right and then just set the cluster behavior of your choice. 
2. Yes it is still recommended since if you don't have fail over servers set up and the client fails to communicate to it's primary node and the cached topology is invalid or non existing  you will only have those fail servers to fall to.
3. I think it is safest to have it set to Replicate, i think this is the default behavior. You can't know who is the leader and you don't want the leader not to propagate changes to other nodes...
in 3.5 we made changes so propagated data will be sent in batches so not to load the network. Generated data will be sent imminently. 
I'm not sure what will happen if you try setting TransitiveReplicationBehavior to None on a destination and then it becomes a leader...
I think it should be set to Replicate which is what i think is the default behavior.

--
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+u...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.



--

Hibernating Rhinos Ltd  cid:image001.png@01CF95E2.8ED1B7D0

Tal Weiss l Core Team Developer Mobile:+972-54-802-4849

Office: +972-4-622-7811 l Fax: +972-153-4-622-7811l Skype: talweiss1982

RavenDB paving the way to "Data Made Simplehttp://ravendb.net/ 

Nick

unread,
Jun 22, 2016, 7:11:25 AM6/22/16
to RavenDB - 2nd generation document database
Thanks Tal.

When you say "then just set the cluster behavior of your choice" I cant see where you would set this?

Tal Weiss

unread,
Jun 22, 2016, 8:42:46 AM6/22/16
to RavenDB - 2nd generation document database
store.Conventions.ClusterBehavior (This is a client side configuration)
Reply all
Reply to author
Forward
0 new messages