Marathon errors
Connection attempt to host-name.net:31022
Symptoms
A user gets the following warning:
WARN Connection attempt to some-host-name.net:31022 failed in response to GET request to /health with 4 retries left, retrying... (the port number can be any number)
Solution
Check the mesos-slaves.ini file on mesos-slaves servers. It must have the ports:\[40000-44000\] port range in the resources parameter. For example:
--resources="cpus:4;mem:15091;ports:\[40000-44000\].
If the parameter is omitted, add it and restart the slave. Here is the place where the parameter should be added. See the screenshot below:
Connection attempt to wfml.host_name.com:42950
Symptoms
A user gets the following warning:
WARN Connection attempt to wfml.host_name.com:42950 failed in response to GET request to /health with 5 retries left, retrying... (the port number can be from range 40000 - 44000)
Solution
It is likely that the port is closed. The 40000-44000 ports used for health checks of the Mesos slave. If all the five checks are failed, Marathon terminates the job. Usually, it looks like the evaluation is being deployed for some time (around 30 minutes) and then crashes.
Deployment failed for change
Symptoms
A user gets the following warning:
WARN Deployment failed for change: 2018-09-25T06:46:34.930Z
(mesosphere.marathon.core.group.impl.GroupManagerActor:marathon-akka.actor.default-dispatcher-15)
mesosphere.marathon.ConflictingChangeException: An app with id
[/extract/flat-prod/—some-mode-name---/2.0.0.0/document-xml-link-71be6837-b914-43ce-8c5d-eda5faf761db]
already exists.
Solution
The application has already been launched on Marathon. You need to wait until it is finished or destroy it if it hangs.