Loading

Verify ZooKeeper sync status

ECE

It is recommended to check the ZooKeeper sync status before starting any maintenance on your Director nodes. This helps you prevent the following scenarios:

  • The ECE UI Settings page displays all ZooKeeper nodes as connected, but not all the nodes have completed the syncing with the latest ZooKeeper state.
  • Connected ZooKeeper nodes participate in the quorum, but they don’t appear in the ECE UI Settings page. For example, if the host is removed, ECE no longer cares about it and keeps the ZooKeeper container part of the quorum.

To check that ZooKeeper is in sync with the correct number of followers, run the following steps:

  1. Run the inline shell script command on each Director node:

    docker exec frc-zookeeper-servers-zookeeper sh -c '
    for i in $(seq 2191 2199); do
      output=$(echo mntr | curl -s telnet://localhost:$i | grep -E "server_state|leader|follower|not currently serving|zk_znode_count");
      if [ -n "$output" ]; then
        echo "ZK mntr Response from port $i:";
        echo "$output";
        break;
      fi
    done'
    
    Note

    You must check all nodes to find the Leader node with the required sync information. This is currently tested only in Bash.

  2. From the Leader node’s output, make sure to check that:

    • The count of followers is correct and expected
    • All followers are listed as synced

The inline shell script command can return the following types of output:

  • If the host is the current ZooKeeper Leader, the command returns the Leader’s info including follower count and follower sync status.

    ZK mntr Response from port 2191:
    zk_server_state leader
    zk_znode_count  783
    zk_synced_followers     2
    ...
    
  • If the host is a follower, the command returns only the follower state, and continues until it finds the Leader:

    ZK mntr Response from port 2193:
    zk_server_state follower
    zk_znode_count  777
    ...
    
  • If the ZooKeeper container is up and listening, but the current node doesn’t have the quorum, the command returns the message This ZooKeeper instance is not currently serving requests:

    ZK mntr Response from port 2192:
    This ZooKeeper instance is not currently serving requests
    

Make sure the ZooKeeper container is running on all the Director nodes. If another Director node is under maintenance, check that ZooKeeper is healthy and synced before starting any other nodes. If all expected nodes are up and running, there might be another issue. Reach out to Elastic support.

If there is no response on any port, it’s possible that no ZooKeeper ports are currently listening (for ex. running on a non-Director role host, or the ZooKeeper Docker container is not running)

If the inline shell script command doesn’t work, you can run the check directly from the director host. This can happen for example when your user lacks permissions to access Docker. This approach avoids entering the container and doesn't require installing additional tools like telnet or nc, relying instead on curl, which is typically available by default on most Linux systems.

  1. Run the equivalent inline shell script directly on the host terminal, outside of the zookeeper container
    for i in $(seq 2191 2199); do
      output=$(echo mntr | curl -s telnet://localhost:$i | grep -E "server_state|leader|follower|not currently serving|zk_znode_count");
      if [ -n "$output" ]; then
        echo "ZK mntr Response from port $i:";
        echo "$output";
        break;
      fi
    done
    
  2. Look for the following lines in the output
  • zk_server_state leader or zk_server_state follower — indicates the node’s ZooKeeper role