None of the enabled containerizers (docker,mesos) could create a container for the provided TaskInfo/ExecutorInfo message.

[Frank Scholten <fr...@frankscholten.nl>]

Hi,

For educational purposes I am setting up a Mesos and Marathon cluster
running in Docker. My goal is to deploy Docker containers on Marathon
using the docker-outside-of-docker technique (DooD) by sharing the
docker command line binary and UNIX socket via a volume.

See the following Gist with a fig.yml file, deploy script and Marathon config

https://gist.github.com/frankscholten/c88efe7d887ee4602138

When I deploy a Docker container via Marathon's API I get task failed
errors in the Mesos GUI. Mesos says that the slave host or port might
not be available.

When I enter a slave using docker exec I am able to run Docker commands.

I checked out the mesos sources at slave/containerizer/docker.cpp and
I see the create method returns Error objects. Do they get logged
somewhere? I the logs I only see

slave_1     | E0115 20:45:21.108027    13 slave.cpp:2794] Container
'27268822-af76-4dac-bf46-feaa60338a7b' for executor
'docker.6c030613-9cf7-11e4-9af3-ee4f42e8b470' of framework
'20150115-203007-301994412-5050-1-0000' failed to start: None of the
enabled containerizers (docker,mesos) could create a container for the
provided TaskInfo/ExecutorInfo message.

so I don't know what the problem is.

Cheers,

Frank

Re: None of the enabled containerizers (docker,mesos) could create a container for the provided TaskInfo/ExecutorInfo message.

[Frank Scholten <fr...@frankscholten.nl>]

Thanks Adam!

Also I had an old version of Marathon. When I switched to 0.7.6 using
the gregory90/marathon image I could start a Docker container.

The only remaining problem is that the Mesos GUI shows the error about
the slave not being available. This is because it performs the slave
state callback using the container ID as the hostname, which does not
resolve.

On Wed, Jan 21, 2015 at 8:33 AM, Adam Bordelon <ad...@mesosphere.io> wrote:
> Found a likely problem in the fig.yml:
> "- MESOS_ISOLATOR=cgroups/cpu,groups/mem"
> That second part should be "cgroups", not "groups"

Re: None of the enabled containerizers (docker,mesos) could create a container for the provided TaskInfo/ExecutorInfo message.

[Adam Bordelon <ad...@mesosphere.io>]

Found a likely problem in the fig.yml:
"- MESOS_ISOLATOR=cgroups/cpu,groups/mem"
That second part should be "cgroups", not "groups"

Re: None of the enabled containerizers (docker,mesos) could create a container for the provided TaskInfo/ExecutorInfo message.

[Tim Chen <ti...@mesosphere.io>]

What is the command you use to run your slave in a Docker container? And
also what version of Mesos are you using?

Tim

On Thu, Jan 15, 2015 at 1:05 PM, Frank Scholten <fr...@frankscholten.nl>
wrote:

> Hi,
>
> For educational purposes I am setting up a Mesos and Marathon cluster
> running in Docker. My goal is to deploy Docker containers on Marathon
> using the docker-outside-of-docker technique (DooD) by sharing the
> docker command line binary and UNIX socket via a volume.
>
> See the following Gist with a fig.yml file, deploy script and Marathon
> config
>
> https://gist.github.com/frankscholten/c88efe7d887ee4602138
>
> When I deploy a Docker container via Marathon's API I get task failed
> errors in the Mesos GUI. Mesos says that the slave host or port might
> not be available.
>
> When I enter a slave using docker exec I am able to run Docker commands.
>
> I checked out the mesos sources at slave/containerizer/docker.cpp and
> I see the create method returns Error objects. Do they get logged
> somewhere? I the logs I only see
>
> slave_1     | E0115 20:45:21.108027    13 slave.cpp:2794] Container
> '27268822-af76-4dac-bf46-feaa60338a7b' for executor
> 'docker.6c030613-9cf7-11e4-9af3-ee4f42e8b470' of framework
> '20150115-203007-301994412-5050-1-0000' failed to start: None of the
> enabled containerizers (docker,mesos) could create a container for the
> provided TaskInfo/ExecutorInfo message.
>
> so I don't know what the problem is.
>
> Cheers,
>
> Frank
>