CODEWITHSUNDEEP
mesos
auxdx
auxdx

Reputation: 2493

Completely Delete orphan task Mesos

I have tried almost every things to delete old mesos tasks but without any success. I have three servers and have installed zookeeper on all of them as well as running mesos-master and mesos-slave in all there servers.

I installed mesos-kafka: https://github.com/mesos/kafka and run kafka-mesos-scheduler is one servers; I then run a broker in each one of the server. For some reason, I messed up with my setup so I have deleted all of the broker. But somehow, after restarting mesos I keep getting something like:

Ignoring status update TASK_LOST (UUID: c223a35d-7807-43af-ac14-570b99729fe3) for task broker-1-181b2793-fb8b-430d-9f26-71a3346a4ee7 of framework 20151009-155844-50393098-5050-19170-0000 from slave 20151009-155844-50393098-5050-19170-S112 at slave(1)@10.240.0.5:5051 (zoo3) because the framework is unknown

although I have tried to stop all of the mesos-master, slave; delete log dir and even stop and delete zookeeper log. But somehow, it still show orphan tasks in log or when i querying the mesos master/state.json. Can anyone please enlighten me how I can completely remove orphan tasks from mesos itself so that I can rollback to a fresh install. It might be because of that, I was no longer able to readd/run mesos kafka broker on those servers again.

Thanks.

Upvotes: 1

Views: 1402

Answers (1)

serejja
serejja

Reputation: 23871

When you run the mesos-kafka scheduler you specify a --storage flag. If you don't, it defaults to a file (kafka-mesos.json). To reset cluster state you should delete your storage znode if you use zookeeper for cluster storage or delete that file. After this starting the scheduler again should have an empty cluster.

Upvotes: 0

Related Questions