I'm reading MESOS-6010 [1] which seems very similar to the problem I'm having but I do have the variables in all lower case and I'm assuming that fix is already part of 1.3.0...
[1] https://issues.apache.org/jira/browse/MESOS-6010 On Thu, Jul 20, 2017 at 3:08 PM, William Markito Oliveira < william.mark...@gmail.com> wrote: > Hi folks, > > I'm trying to setup a Mesos cluster and submit the "hello world" GPU > example from the documentation page: > > *"mesos-execute --master=10.120.59.5:5050 <http://10.120.59.5:5050> > --name=gpu-test --docker_image="nvidia/cuda" > --command="nvidia-smi" --framework_capabilities="GPU_RESOURCES" > --resources="gpus:1"* > > > This returns: > > I0720 16:02:26.039414 102623 scheduler.cpp:184] Version: 1.3.0 > I0720 16:02:26.040221 102619 scheduler.cpp:470] New master detected at > master@10.120.59.5:5050 > Subscribed with ID 4d3b5156-85df-4ffd-a8cd-9e0ecaa90e39-0015 > Submitted task 'gpu-test' to agent 'b2d906e8-1207-4ceb-aeb0- > 42be1150cff8-S4' > Received status update TASK_FAILED for task 'gpu-test' > message: 'Failed to launch container: Failed to perform 'curl': curl: > (7) Failed to connect to registry-1.docker.io port 443: Connection refused > ' > source: SOURCE_AGENT > reason: REASON_CONTAINER_LAUNCH_FAILED > > --- > > Now when specifying the containerizer=docker I receive the following > output: > > mesos-execute --containerizer=docker --master=10.120.59.5:5050 > --name=gpu-test --docker_image="nvidia/cuda" > --command="nvidia-smi" --framework_capabilities="GPU_RESOURCES" > --resources="gpus:1" > I0720 16:02:59.589102 102719 scheduler.cpp:184] Version: 1.3.0 > I0720 16:02:59.589792 102727 scheduler.cpp:470] New master detected at > master@10.120.59.5:5050 > Subscribed with ID 4d3b5156-85df-4ffd-a8cd-9e0ecaa90e39-0016 > Submitted task 'gpu-test' to agent 'b2d906e8-1207-4ceb-aeb0- > 42be1150cff8-S3' > Received status update TASK_RUNNING for task 'gpu-test' > source: SOURCE_EXECUTOR > Received status update TASK_FAILED for task 'gpu-test' > message: 'Container exited with status 127' > source: SOURCE_EXECUTOR > > So still no success, but a different error. > > My environment does have http_proxy and https_proxy variables with proper > values and I've set them before starting the agents. > > Both docker and nvidia-docker pull works just fine and can download the > images. > > Any thoughts on how to fix or troubleshoot this ? > > Thank you! > > Version: Mesos 1.3.0 > OS: Ubuntu 16.04 > Docker: 17.06.0-ce (+ NVIDIA-Docker) > > > -- > ~/William >
