Open swb2-izu-ssp opened 3 days ago
This will be a side effect from the way that cluster discovery works, the error would still show in the logs but would fade out after some time. The important part is to make sure that all the nodes have the same extra routes to avoid partitions, so if you change the configuration maps from both clusters to include the explicit routes and then issue a config reload you would have the mesh being formed.
Hello @wallyqs Many thanks for your answer.
The configuration map of the 3 parts looks like this
PART 1
routes:
- tls://stretch-cluster-0.nats.nats-stretch-1.svc.cluster.local:6222
- tls://stretch-cluster-1.nats.nats-stretch-1.svc.cluster.local:6222
- tls://stretch-cluster-2.nats.nats-stretch-1.svc.cluster.local:6222
- tls://nats-nats-stretch-2.mydomain.com:443
- tls://nats-nats-stretch-3..mydomain.com:443
PART 2
routes:
- tls://stretch-cluster-0.nats.nats-stretch-2.svc.cluster.local:6222
- tls://stretch-cluster-1.nats.nats-stretch-2.svc.cluster.local:6222
- tls://stretch-cluster-2.nats.nats-stretch-2.svc.cluster.local:6222
- tls://nats-nats-stretch-1.mydomain.com:443
- tls://nats-nats-stretch-3..mydomain.com:443
PART 3
routes:
- tls://stretch-cluster-0.nats.nats-stretch-3.svc.cluster.local:6222
- tls://stretch-cluster-1.nats.nats-stretch-3.svc.cluster.local:6222
- tls://stretch-cluster-2.nats.nats-stretch-3.svc.cluster.local:6222
- tls://nats-nats-stretch-1.mydomain.com:443
- tls://nats-nats-stretch-2..mydomain.com:443
As a matter of fact, I made it works, by disabling the no_adveritise key. So, in a configuration, where I am deploying the 3 namespaces on same paas, this works.
But as soon, as I deployed the 3 namespaces on 3 different paases, I now have cluster un-stabilities.
This is varying if I run several time the nats list server cmd. What I am missing?
Nicolas
Observed behavior
Hello,
I am trying to put in place the following pattern: Stretch cluster, coming from https://www.synadia.com/blog/multi-cluster-consistency-models/
To do so, servers re-unification dispatched over several paases must be done within a single cluster.
The solution I have found so far is to use openshift route.
What I have tried before using route, is to deploy servers on different namespaces, but same paas, and use the headless service only for the route urls.
it is working fine.
So I tried to extend this with openshift routes (with tls termination passthrough + routing to port 6222) only for the remote connectivity (locally, I am using Pod+headless service) and still same configuration (different namespaces and same paas)
I started having issue with TLS because it is trying to connect with IP at the end, and not the hostname I have provider
So it seems, with route, it is not passing tlsName for tls validation. And of course, my certificate does not contain Ip adresses. I would suspect this piece of code: https://github.com/nats-io/nats-server/blob/main/server/client.go#L5905 But this is a fast and lazy check, I have done...
Expected behavior
I would expect to see it works also, same as headless service usage.
Server and client version
Service: 2.10.21
Host environment
Linux
Steps to reproduce
No response