After Add Hosts to Nebula 2.0,the leader metad instance will have different meta data when scheduled to different nodes

  • Nebula version:2.0.1
  • Deployment:K8S
    Description:
    I deploy a Nebula cluster with k8s,after a series operations like snapshot create,space create,space drop,host add(when I add host, the graphd,metad,storaged instance count changes from 3 to 5),data balance, and I found a fatal problem.
    When I restart the cluster,the leader metad instance will be scheduled to different nodes, and it will have different meta data when it was scheduled to the old nodes and the new added nodes.For example, when the leader metad instance is shceduled to the old nodes,it will have a space named A with space id 1,then when I restart the cluster and the leader metad instanct is shceduled to the new nodes, it will have a space named B with space id 1 as well. And both space A and B share the storage data path under the directory ./data/storage/nebula/1/
    , both space have data but don’t have the same amount.I have no idea how can this problem occur.

@turtlebin Thanks for posting your question here! I am sure your problem has been discussed in detail here: After Add Hosts to Nebula 2.0,the leader metad instance will have different meta data when scheduled to different nodes · Issue #1041 · vesoft-inc/nebula-graph · GitHub
Please update the GitHub issue if there is anything new to add.