Scheduler should remove its frameworkId from zookeeper on shutdown
epeld opened this issue · comments
Currently it is not possible to restart a scheduler once it has terminated because the path /logstash/frameworkId already exists in zookeper and we get a collision.
The scheduler should make sure it erases its frameworkId as soon as it terminates so that it can be restarted.
The ZK path /logstash/frameworkId represents the ID of the single Logstash framework registered with Mesos. The reason we store this is so that we can restart a scheduler under the same framework ID. The current behavior is that, when the scheduler starts, if a value exists at /logstash/frameworkId, the scheduler uses that value as its framework ID when registering with Mesos.
Therefore I don't think we have any collision problems, and this issue has been fixed.