0

New to Ray, trying to do some troubleshooting on distribution of tasks across the cluster.

Right now we have the head node doing all of the work, and not delegating anything at all to workers. Hoping to get some suggestions on logs to investigate to track down potential errors. How / where can I see the tasks that a worker node is processing?

jsleeuw
  • 283
  • 5
  • 13
  • Make sure to check `ray.nodes()` to verify that all of the nodes are part of the cluster. You can also try the dashboard to see which workers are active. You can start the dashboard by running `ray start --head --include-webui` (you may want to try the nightly wheels to get the latest version of the dashboard, see https://ray.readthedocs.io/en/latest/installation.html#latest-snapshots-nightlies). Another thing to try is the Ray timeline (run `ray timeline` from the command line while Ray is running). – Robert Nishihara Oct 26 '19 at 06:15
  • Thanks Robert, really appreciate the pointers. I'll get back on the thread here after I've had a chance to try out your suggestions. – jsleeuw Oct 26 '19 at 13:13
  • 1
    OK, I got this working by explicitly passing `redis_address` to ray.init(). I must have picked up an example script somewhere that had that blank. – jsleeuw Oct 28 '19 at 20:34

0 Answers0