apache/accumulo

accumulo-cluster does not start a CompactionCoordinator on different host

ddanielr opened this issue · 2 comments

ddanielr commented

Describe the bug
The accumulo-cluster script does not start a compaction-coordinator process if the defined host is different than the local manager node.

SSHing to the expected node, an accumulo-cluster start-here command also failed to start the coordinator.

Versions (OS, Maven, Java, and others, as appropriate):

  • Affected version(s) of this project: [e.g. 1.10.0] 2.1.x
  • OS: [e.g. CentOS 7.5] centos 7
  • Others:

To Reproduce

  1. Modify the cluster.yaml to have the compaction-coordinator service defined on a host other than localhost and the hostname used for the gc, monitor, and manager.
  2. Run an accumulo-cluster start command
  3. SSH to the expected host and ensure no process is running.
  4. Make sure the cluster.yaml is updated on that host then run ./accumulo-cluster start-here.
  5. Verify that the process did not attempt to start by looking at the log files or pid files in accumulo's run directory.
  6. Finally, run a accumulo-service compaction-coordinator start operation and verify the process attempted to start and is running.

Expected behavior
The cluster script is able to start and stop the compaction-coordinator no matter what host it is running on.

dlmarion commented

I was unable to replicate this. I modified accumulo-cluster locally to emit debug output and so that it would always take the ssh path. I set the hosts in cluster.yaml to the public IP address of my dev box, but set my manager address to localhost. I saw the ssh command come across and compaction-coordinator logs appeared in the logs directory (Accumulo wouldn't start because I had ZK and Hadoop down).

ddanielr commented

Closing this as unable to reproduce