NSX Manager is down or unavailable if majority of the nodes in the cluster are down.

Problem

NSX Manager UI will fail to load with following error Some appliance components are not functioning properly. Component health: POLICY:UNKNOWN, MANAGER:UNKNOWN, SEARCH:UNKNOWN, NO and clustering related commands will fail using the CLI and API.

Solution

  1. SSH to each of the affected NSX Manager nodes and run following commands:
    1. Run get file-system-stats and verify /config and /image is not 100% full.
    2. Run get core-dumps to verify no cores have gotten generated in NSX Manager.
    3. Verify there was no datastore outage. See NSX Manager Cluster status Degraded As Datastore-related Components Are Down.
    4. Check logs for out-of-memory errors. See /var/log/proton/proton-tomcat-wrapper.log
  2. To restore clustering and UI, any two nodes in a three node cluster must be up. If you are not able to bring any failed node back up, but if there is a healthy node available, then do one of the following steps to restore clustering:
    • Deploy a new manager node (as 4th member node), join the existing cluster and then detach one of the failed nodes using CLI cmd detach node <node-uuid> or API POST /api/v1/cluster/<node-uuid>?action=remove_node. The commands should be executed from one of the healthy nodes. Alternatively, you can follow the next bulleted point to deactivate the cluster.
    • (Optional) Run the deactivate cluster command on active node such that you end up with single node cluster. Now continue to add the new additional nodes to make a 3-member NSX Manager cluster.
      Note: NSX Manager nodes that are removed from the cluster should be powered off and deleted.