Failover for Neo4j Causal Cluster might be easier than what you think.


Use bolt+routing protocol for your neo4j client driver to connect to Neo4j causal cluster.
(If the Neo4j driver you are using supports Bolt Routing Protocol)


I've been using Neo4j about two months. I eastablished a Neo4j causal cluster via Helm chart of Neo4j and used it in production with 3 core nodes (1 Leader, 2 Followers) and 3 read replicas.


Recently, I got an error like this: "Not A Leader Error: Can't write to FOLLOWER node". It always occured after the leader node became follower node. That's the time I realized there might be a problem of the failover process.


At first, I was trying hard to find a way to make the Kubernetes service of my Neo4j causal cluster to always redirect conneciton to leader node. Because I know that Neo4j causal cluster comes up with default service discovery mechanism, I thought maybe add a sidecar container to watch the result of cypher shell dbms.cluster.overview(); and change the pod label for the pod, so I can add a selector into Kubernetes service to do what I want.

After I read some posts and codes about how to write a sidecar container in a pod, I still don't know how to do. At the time I almost gave up, here comes the salvation:

bolt+routing://<host>:<port> caught my eyes.


It's really simple, but it looks like not many people who use Neo4j causal cluster know about it. I've spent so much time on finding the answer.
Official documentation of Neo4j also mentioned about it: (I should've read official doc more carefully, so I don't have to waste so much time on the wrong path.)
It turns out if you're using bolt protocol (7687 port) for your Neo4j client to connect to Neo4j cluster and the Neo4j client you're using support the Bolt Routing Protocol, you can just simply change the url of Neo4j:

from bolt://<host>:<port>
to bolt+routing://<host>:<port>
(The <host> here for me is the Kubernetes service name of my Neo4j causal cluster)

I am using py2neo as my Neo4j driver and it happens to support Bolt Routing Protocol. HOW LUCKY.

That's it.



如果覺得這篇文章對你有幫助, 除了留言讓我知道外, 或許也可以考慮請我喝杯咖啡, 不論金額多寡我都會非常感激且能鼓勵我繼續寫出對你有幫助的文章。

If this blog post happens to be helpful to you, besides of leaving a reply, you may consider buy me a cup of coffee to support me. It would help me write more articles helpful to you in the future and I would really appreciate it.