Currently I’ve configured cluster to use hostnames.
Everything is ok and all clients have access to cluster.
And nodes communicate with each other by hostnames - by public network. And it’s slower than private network.
If I configure cluster using private IP’s, clients that’re not in private network will not be able to access the cluster.
How should I configure cluster to make nodes communicate with each other by private network and allow client from public network access it?
This is definitely an issue we’re aware of and working on that will be available in a future release. It’s come up in the context of various cloud providers, but the improvement should also be able to help you out.
I have actually used the internal comms for adding hosts to the cluster and it seems that both internal/external communication is taken care of. I think using DNS/ host files would have complicated a matter as you would not know which ones are used by the system /etc and we have got no DNS available on this system.
I would like to know if this is the supported configuration on the couchbase system.
I have got Eth0 - public IP -< available externally
and Eth1 - private IP (RFC1918) -< used for adding hosts into the cluster and for private comms.
I’m working getting the details (and a case number if I can find one), but I think there plan is to support it in 5.5. (I’m not sure if it’s supported in the current 5.5 developer build that’s available for download).
As @matthew.groves mentioned we shipped a server feature “alternate address” in Couchbase Server 5.5.0 release here is the JIRA MB-24250 but this feature is not fully GA with documentation until (next release) upcoming 6.5.0 release. Here is the JIRA MB-32751DOC-4493 for GA in 6.5.0.
Hi Anil, I tried but could not access either of the docs/pages.
Meanwhile, as I understand from your response, at this point it’s not possible. Even if all of the mentioned 4 are not possible, is any 1/2/3 possible?
Alternately, any workarounds that you can suggest? In enterprise environments, as you would be aware, these are different subnets/vLANs. And this should be a common req so other users would be working around somehow, I guess.