Seeding a Causal Cluster on Kubernetes

I'd like to start a discussion on the most performant way to seed a causal cluster running on Kubernetes.

The starting point of my seeding process is I have a graph.dump file with the database I'd like to load on each core/replica instance. I believe the next step would be to run init containers that copy that dump file to their respective persistent volume, and then run the neo4j-admin load tool to import that database to graph.db and finish the init container. At this point the Causal cluster will start up and the core members will discover eachother and everything would be grand.

Does this seem like the right process? Would it be possible to use a shared volume for the graph.dump file and then all persistent volumes can be loaded off of that one file? Let me know if there are any fatal flaws in my process please!

1 Like

That does seem like the right process. That is in fact how I recommended that you do it, except that I was using a neo4j backup proper rather than a dump file.

There's a full article writeup on this technique and how to do it here:

But for short -- I think you've got the general idea correct, and that will work.

1 Like