Mesos frameworks register but tasks will not start

Symptoms:

  • Frameworks register in Mesos, but tasks will not start.
  • The Frameworks tab of the Mesos Master UI shows frameworks registered on 'localhost'.
  • mesos-fetcher logs report the following:
    Scheduler driver bound to loopback interface! Cannot communicate with remote master(s). You might want to set 'LIBPROCESS_IP' environment variable to use a routable IP address.

Resolution:

  1. Confirm that each node correctly resolves its own hostname.
  2. Ensure the following line exists in /etc/hosts (substituting x.x.x.x for the host's resolvable IP address, and somehostname for the hostname):
    x.x.x.x somehostname
  3. Uninstall the framework:
    dcos package uninstall someframework
  4. Ensure that the framework is removed from the Mesos Master UI (replace $FRAMEWORK_ID with the correct framework ID):
    echo "frameworkId=$FRAMEWORK_ID" | curl -X POST -d@- http://mesos.master.com:5050/master/shutdown
  5. Remove any Zookeeper ZNodes related to the framework.
  6. Re-install the framework:
    dcos package install someframework

 

Have more questions? Submit a request

Comments

Powered by Zendesk