milvus operatorIn a Milvus cluster deployed by the Milvus operator, the internal etcd cluster has a pod whose physical machine fails. How do I migrate this pod?

[bozai111]

In a Milvus cluster deployed by the Milvus operator, the internal etcd cluster has a pod whose physical machine fails. How do I migrate this pod?

Does the Milvus Operator support automatic migration of the built-in etcd pod in failure scenarios?

If automatic migration is not supported, how do I perform manual migration?

Is there more information available about the Milvus Operator?

[haorenfsa]

If you use the the kubernetes provided by Cloud Providers, its PVC is a cloud disk can be automatically moved to other pod in the same availability zone. so you just need to delete your old pod, and it get scheduled to another machine, and the cloud disk volume will also be mounted to the new machine.

If you're using a self build kubernetes, we suggest you you use a portable PVC provisioner like ceph, so that you can use the same mechanism.

[bozai111]

The current etcd pod uses a local disk for LVM. The machine in the etcd cluster has failed and cannot be started. How should I migrate etcd?

[bozai111]

The performance of Ceph is much worse than that of local disks. The official documentation of Milvus recommends that etcd be placed on a high-performance disk of NVMe.