Andrei Budnik created MESOS-10098:
-------------------------------------

             Summary: Mesos agent fails to start on outdated systemd.
                 Key: MESOS-10098
                 URL: https://issues.apache.org/jira/browse/MESOS-10098
             Project: Mesos
          Issue Type: Bug
          Components: agent
    Affects Versions: 1.10
         Environment: CoreOS 2411.0.0
            Reporter: Andrei Budnik
            Assignee: Andrei Budnik
             Fix For: 1.10


Mesos agent refuses to start due to a failure caused by the systemd-specific 
code:
{code:java}
E0220 12:03:02.943467 22298 main.cpp:670] EXIT with status 1: Expected exactly 
one socket with name unknown, got 0 instead
{code}

It turns out that some versions of systemd do not set environment variables 
`LISTEN_PID`, `LISTEN_FDS` and `LISTEN_FDNAMES` to the Mesos agent process, if 
its systemd unit is ill-formed. If this happens, `listenFdsWithName` returns an 
empty list, therefore leading to the error above.

After fixing the problem with the systemd unit, systemd sets the value for 
`LISTEN_FDNAMES` taken from the `FileDescriptorName` field. In our case, the 
env variable is set to `systemd:dcos-mesos-slave`. Since the value is expected 
to be equal to "systemd:unknown" (for the compatibility with older systemd 
versions), the mismatch of values happens and we see the same error message.

 



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

Reply via email to