High Availability Configuration of Conductor Server about conductor

[v1r3n]

Yes, we do actually run it that way. Conductor servers can scale up by adding more servers and sharing the same instance of Dynomite.

[blueelephants]

Thanks for clarification

[cquon]

@v1r3n a quick question on this. Is the scaling stateless or is it more of a cluster? Is there any notion of primary/slaves or all work as standalone servers with same database.

Also is there any special handling that needs to be taken for potential race conditions with a scaled architecture?

[v1r3n]

@cquon the scaling is stateless and follows peer to peer architecture and not master/slave. Each server in the cluster is capable of serving the requests. The cluster is identified by the group of servers talking to the same underlying Dynomite database.

Conductor takes care of any race conditions so no special care has to be taken.

[akashbhunchal]

Is this applicable even for redis (not dynomite) ? I am facing race conditions when I try to run more than conductor servers with one redis instance as the datastore. I have raised an issue with my observations on synchronized block in the ReqisQueue implemetation of pop()

#543