[ 
https://issues.apache.org/jira/browse/CASSANDRA-11107?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15706743#comment-15706743
 ] 

Joel Knighton commented on CASSANDRA-11107:
-------------------------------------------

I've got a patch in progress that solves the easy parts of this. At this point, 
however, I am having second thoughts regarding the costs/benefits of this 
change.

At this point, to support separate rpc/native_transport configurations, changes 
would seem to include:
* updating the native protocol so that NEW_NODE events include rpc_address and 
native_transport_address (and other TopologyChangeEvents, since identifiers 
used by drivers might include both address configurations)
* updating the PEERS table to include rpc_address and native_transport_address
* adding an ApplicationState in Gossip for native_transport_address.

Drivers would also need to be updated to query native_transport_address 
appropriately. This seems like a fair amount of work when 4.0 will end up 
negating these changes on removing Thrift.

The other option that immediately presents itself is to allow these properties 
to be set in a 3.X yaml but require them to match the rpc configurations. I'm 
not sure this is worth it either.

Let me know what you think, [~slebresne].

> Add native_transport_address and native_transport_broadcast_address yaml 
> options
> --------------------------------------------------------------------------------
>
>                 Key: CASSANDRA-11107
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-11107
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Configuration
>            Reporter: n0rad
>            Assignee: Joel Knighton
>            Priority: Minor
>
> I'm starting cassandra on a container with this /etc/hosts
> {quote}
> 127.0.0.1        rkt-235c219a-f0dc-4958-9e03-5afe2581bbe1 localhost
> ::1          rkt-235c219a-f0dc-4958-9e03-5afe2581bbe1 localhost
> {quote}
> I have the default configuration except :
> {quote}
>  - seeds: "10.1.1.1"
> listen_address : 10.1.1.1
> {quote}
> cassandra will start listening on *127.0.0.1:9042*
> if I set *rpc_address:10.1.1.1* , even if *start_rpc: false*, cassandra will 
> listen on 10.1.1.1
> Since rpc is not started, I assumed that *rpc_address* and 
> *broadcast_rpc_address* will be ignored
> It took me a while to figure that. There may be something to do around this



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to