In Cassandra3 storage, only use consistency level LOCAL_ONE if a local_dc has been defined, otherwise use ONE #1753
Conversation
…l_dc has been defined, otherwise use ONE. Mark all writes as idempotent. Update the retry policy, honouring idempotence, consistency level, and applying appropriate back-pressure.
There was a problem hiding this comment.
reformatted w/ https://github.com/square/java-code-styles vs nagging you :)
just one thing on my end, just to understand this. @llinder how do you feel?
| @@ -123,6 +125,13 @@ static Cluster buildCluster(Cassandra3Storage cassandra) { | |||
| builder.withPoolingOptions(new PoolingOptions().setMaxConnectionsPerHost( | |||
| HostDistance.LOCAL, cassandra.maxConnections | |||
| )); | |||
|
|
|||
| builder.withQueryOptions( | |||
There was a problem hiding this comment.
this is the only one I don't know just by reading it. Can you put a comment line of why we do this, or tell me and I'll add it? Thx!
There was a problem hiding this comment.
if local_dc isn't defined then LOCAL_ONE sticks requests to the dc of the first seed host that the driver connects to.
ONE will at least in this situation connect to any dc, and then any latency awareness will bring it back to (eventually) using the local dc.
furthermore LOCAL_ONE can fail easily if you're not using 'NetworkTopologyStrategy` but that's a fix for another day.
There was a problem hiding this comment.
added comment. (can you please squash the 4 commits!)
|
thanks tons!
|
…l_dc has been defined, otherwise use ONE (openzipkin#1753) Mark all writes as idempotent. Update the retry policy, honouring idempotence, consistency level, and applying appropriate back-pressure.
…l_dc has been defined, otherwise use ONE (openzipkin#1753) Mark all writes as idempotent. Update the retry policy, honouring idempotence, consistency level, and applying appropriate back-pressure.
Also…