scylladb
[question] Kubernetes - ScyllaDB
I was taking a look at how ScyllaDB handled integration with Kubernetes and I came upon this solution. While it is very elegant and utilizes Kubernetes' built-in primitives, I have a few questions, coming from a Cassandra background:
- How do you handle Node / Disk failure ? Cassandra requires that you supply the IP address of the dead instance and I imagine that Scylla requires that too.
- Is there any mechanism for backup/restore/repair of the cluster without having to visit each node separately ?
- What are other limitations you have encountered with this solution that could be tackled by developing a Kubernetes operator ? Do you find that it needs manual intervention often ?
@hakuch can you help?
BTW, a better place for such a question will be Scylla user list https://groups.google.com/d/forum/scylladb-users
As I've written in the blog post, a production deployment of Scylla with Kubernetes will need to address many of the issues you're describing and this is beyond the scope of this code sample and introduction.
I think these are excellent conversation topics for a discussion on the users mailing list (as Tzach has already referenced).
