Casey Sybrandy created MESOS-3745:
-------------------------------------
Summary: Slaves makes port unavailable even though no processes
are using/blocking it.
Key: MESOS-3745
URL: https://issues.apache.org/jira/browse/MESOS-3745
Project: Mesos
Issue Type: Bug
Components: slave
Affects Versions: 0.24.1
Reporter: Casey Sybrandy
Hello,
I'll try to describe this as best I can. I was trying to get a service running
on some nodes and it would fail on about half of them. The logs on the master
would state that port 1234 was not part of the offer. I looked at the nodes
and could not find any processes using that port. I manually ran the docker
container I was trying to start on one of the nodes and it worked fine. I
ended up stopping the slave, removing it's data, then starting it back up to
resolve it. Unfortunately, I don't know why this occurred.
Ideally, this shouldn't be happening, however it's understandable that
unexpected events could occur that may put the slave in a weird state. Perhaps
a utility could be created that could list the resource usage information on
that slave from the slave's point of view and provide the ability to free up
resources without restarting the slave? Being a dev system this was fine, but
I'd hate to have to restart a slave on a production system if I can help it.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
