Hi Krish,

>From the output of mesos-execute, seems the framework did not start to
register, does the IP address of mesos master is right? Can you please run
the following command to see if Mesos master is running well?

curl "http://10.20.3.100:5050/master/state.json"; 2>/dev/null| jq .

I also did some test with Mesos 0.24.1 and it works well with docker
container. My steps are as following:

1) Get mesos source code and switch to 0.24.1, build and get all mesos
binaries.
2) Start up mesos master:

root@mesos002:~/src/mesos/m1/mesos/build# ./bin/mesos-master.sh
--work_dir=/var/lib/mesos

3) Start mesos slave with "docker" as containizer:

root@mesos002:~/src/mesos/m1/mesos/build# ./bin/mesos-slave.sh --master=
192.168.56.12:5050  --resources="cpus:8;mem:8000" --containerizers="docker"
--executor_registration_timeout=5mins --log_dir=/opt/mlog/

4) Submit a task with mesos-execute:

root@mesos002:~/src/mesos/m1/mesos/build# ./src/mesos-execute --master=
192.168.56.12:5050  --command="sleep 10"  --name=test_mesos
--docker_image=busybox:latest

I0228 20:29:43.748414 26684 sched.cpp:164] Version: 0.24.1

I0228 20:29:43.753610 26700 sched.cpp:262] New master detected at
master@192.168.56.12:5050

I0228 20:29:43.754464 26700 sched.cpp:272] No credentials provided.
Attempting to register without authentication

I0228 20:29:43.758775 26698 sched.cpp:640] Framework registered with
20160228-202729-205039808-5050-26440-0001

Framework registered with 20160228-202729-205039808-5050-26440-0001

task test_mesos submitted to slave 20160228-202729-205039808-5050-26440-S0

Received status update TASK_RUNNING for task test_mesos

Received status update TASK_FINISHED for task test_mesos

I0228 20:29:54.334636 26698 sched.cpp:1748] Asked to stop the driver

I0228 20:29:54.334771 26698 sched.cpp:1032] Stopping framework
'20160228-202729-205039808-5050-26440-0001'

5) When task is running, check docker container:

root@mesos002:/home/gyliu# docker ps

CONTAINER ID        IMAGE               COMMAND                  CREATED
          STATUS              PORTS               NAMES

e6f08a3661e7        busybox:latest      "/bin/sh -c 'sleep 10"   10 seconds
ago      Up 9 seconds
mesos-20160228-202729-205039808-5050-26440-S0.bab71e31-8de9-44a1-9f5c-8c46ebeedb34

6) When task is finished, the docker container will be stopped.

root@mesos002:/home/gyliu# docker ps

CONTAINER ID        IMAGE               COMMAND             CREATED
    STATUS              PORTS               NAMES

Thanks,

Guangya

On Sun, Feb 28, 2016 at 5:39 PM, Krish <krishnan.k.i...@gmail.com> wrote:

> This is the command I run and it just hangs there - no registration with
> mesos master.
>
> I had to create a /etc/hosts file inside my mesos agent container to get
> rid of the message: "Failed to obtain the IP address for 'minion'; the DNS
> service may not be able to resolve it: Name or service not known"
>
> root@minion:/# mesos-execute --master=10.20.3.100:5050  --command="sleep
> 100"  --name=test_mesos --docker_image=repo.example.com:5000/krish/test
>
> I0228 04:56:23.078397 13778 sched.cpp:164] Version: 0.24.1
> I0228 04:56:23.080368 13786 sched.cpp:262] New master detected at
> master@10.20.3.100:5050
> I0228 04:56:23.080498 13786 sched.cpp:272] No credentials provided.
> Attempting to register without authentication
>
>
>
>
> --
> κρισhναν
>
> On Sun, Feb 28, 2016 at 8:14 AM, Guangya Liu <gyliu...@gmail.com> wrote:
>
>> Hi Krish,
>>
>> Can you please show more how did you do the test? Are you failing when
>> start slave or failed to run mesos-execute?
>>
>> Just check code of 0.24.1, found that the mesos slave does support both 
>> containerized
>> (docker) & non-containerized workloads running at the same time, please
>> refer to
>> https://github.com/apache/mesos/blob/0.24.1/src/slave/flags.cpp#L346
>>
>> But the mesos-execute do not have the option of containerizer in 0.24.1,
>> please refer to
>> https://github.com/apache/mesos/blob/0.24.1/src/cli/execute.cpp#L205-L214
>> , so if you are using 0.24.1 mesos-execute and want to create docker
>> container, there is no need to specify containerizer but only specify
>> the docker_image is enough.
>>
>> Thanks,
>>
>> Guangya
>>
>> On Sun, Feb 28, 2016 at 2:29 AM, Krish <krishnan.k.i...@gmail.com> wrote:
>>
>>> Well, I tried both. But the help that is printed contains no
>>> containerizer  or containerizers field.
>>>
>>>
>>> On Saturday 27 February 2016, <connor....@gmail.com> wrote:
>>>
>>>> The flag name is --containerizers (plural)
>>>>
>>>> HTH,
>>>> --
>>>> Connor
>>>>
>>>> On Feb 27, 2016, at 08:54, Krish <krishnan.k.i...@gmail.com> wrote:
>>>>
>>>> Thanks Guangya for your help.
>>>>
>>>> Unfortunately, something seems to be wrong. I get the error that the
>>>> 'containerizer' is not supported: "Failed to load unknown flag
>>>> 'containerizer'"
>>>> I am using mesos-0.24.1.
>>>>
>>>> I am trying to run a hello-docker aurora (v0.11) job using docker
>>>> containerizer.
>>>>
>>>>
>>>>
>>>> --
>>>> κρισhναν
>>>>
>>>> On Mon, Feb 22, 2016 at 12:08 PM, Guangya Liu <gyliu...@gmail.com>
>>>> wrote:
>>>>
>>>>> I think that the following cli is a good start point for you to take a
>>>>> look which supports all containizers in mesos.
>>>>>
>>>>> https://github.com/apache/mesos/blob/master/src/cli/execute.cpp
>>>>>
>>>>> ./src/mesos-execute --master=192.168.56.12:5050  --command="sleep
>>>>> 1000"  --name=test_mesos --docker_image=busybox:latest --
>>>>> containerizer=mesos
>>>>>
>>>>> ./src/mesos-execute --master=192.168.56.12:5050  --command="sleep
>>>>> 1000"  --name=test_mesos --docker_image=busybox:latest —
>>>>> containerizer=docker
>>>>>
>>>>> Thanks,
>>>>>
>>>>> Guangya
>>>>>
>>>>> On Mon, Feb 22, 2016 at 2:22 PM, Krish <krishnan.k.i...@gmail.com>
>>>>> wrote:
>>>>>
>>>>>> I was wondering if we can have both containerized (docker) &
>>>>>> non-containerized workloads running at the same time on Mesos 
>>>>>> agents/slaves.
>>>>>>
>>>>>> I see from the mesos-slave command help that there is a
>>>>>> 'containerizers' option in the CLI & it will enforce the usage of
>>>>>> containers to be used. It doesn't seem to support non-containerized
>>>>>> workloads.
>>>>>>
>>>>>> Is my analysis correct?
>>>>>>
>>>>>> Also, if the above is correct, how does one tell the frameworks
>>>>>> (Aurora/Marathon/Cassandra frameworks) to use containerized vs
>>>>>> non-containerized workload slaves to schedule jobs?
>>>>>>
>>>>>> --
>>>>>> κρισhναν
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> Guangya Liu (刘光亚)
>>>>> Senior Software Engineer
>>>>> DCOS and OpenStack Development
>>>>> IBM Platform Computing
>>>>> Systems and Technology Group
>>>>>
>>>>
>>>>
>>>
>>> --
>>>
>>> Thumb typed mail
>>>
>>>
>>
>

Reply via email to