[
https://issues.apache.org/jira/browse/MESOS-8065?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ambuj Pandey updated MESOS-8065:
--------------------------------
Description:
I tried changing the "docker-no-executor-framework" to require GPU resource and
use mesos containerizer with a docker image of nvidia/cuda.
This worked for me when i ran this framework with the master IP of one
installed with DC/OS. But when I manually compiled Mesos (1.4/1.5) and started
agent/master on another machine I face this issue : "Implicitly declining
offers in ACCEPT call as the launch operation specified no tasks."
I've tested the nvidia/cuda image to work on that target machine using
nvidia-docker.
Configuration:
Master -
./bin/mesos-master.sh pas-lab-server6:5050 --work_dir=/home/sayantan/mesos_temp
--filter_gpu_resources=false
Agent (on the same node) -
./bin/mesos-agent.sh --master=pas-lab-server6:5050
--work_dir=/home/sayantan/mesos_temp
--isolation="filesystem/linux,cgroups/devices,gpu/nvidia"
I1010 14:48:44.106006 20372 sched.cpp:336] New master detected at
[email protected]:5050
I1010 14:48:44.107053 20372 sched.cpp:351] No credentials provided. Attempting
to register without authentication
I1010 14:48:44.113246 20365 sched.cpp:747] Framework registered with
00afc79a-cb51-4f9d-8c78-6000253ae396-0000
Registered!
.Starting task 0 on pas-lab-server6
Hello
Task 0 is in state 3
. . . . . .
The log on master says:
{code:java}
I1010 14:48:44.420087 19698 master.cpp:7766] Sending 1 offers to framework
00afc79a-cb51-4f9d-8c78-6000253ae396-0000 (Docker No Executor Framework (C++))
at [email protected]:45628
W1010 14:48:44.422901 19701 master.cpp:3911] Implicitly declining offers: [
00afc79a-cb51-4f9d-8c78-6000253ae396-O1 ] in ACCEPT call for framework
00afc79a-cb51-4f9d-8c78-6000253ae396-0000 as the launch operation specified no
tasks
I1010 14:48:44.423539 19701 master.cpp:9389] Removing offer
00afc79a-cb51-4f9d-8c78-6000253ae396-O1
I1010 14:48:44.423607 19701 master.cpp:4196] Processing ACCEPT call for offers:
[ 00afc79a-cb51-4f9d-8c78-6000253ae396-O1 ] on agent
00afc79a-cb51-4f9d-8c78-6000253ae396-S0 at slave(1)@127.0.1.1:5051
(pas-lab-server6) for framework 00afc79a-cb51-4f9d-8c78-6000253ae396-0000
(Docker No Executor Framework (C++)) at
[email protected]:45628
I1010 14:48:44.425109 19691 hierarchical.cpp:887] Updated allocation of
framework 00afc79a-cb51-4f9d-8c78-6000253ae396-0000 on agent
00afc79a-cb51-4f9d-8c78-6000253ae396-S0 from gpus(allocated: *):2;
cpus(allocated: *):40; mem(allocated: *):63421; disk(allocated: *):868675;
ports(allocated: *):[31000-32000] to gpus(allocated: *):2; cpus(allocated:
*):40; mem(allocated: *):63421; disk(allocated: *):868675; ports(allocated:
*):[31000-32000]
{code}
was:
I tried changing the "docker-no-executor-framework" to require GPU resource and
use mesos containerizer with a docker image of nvidia/cuda.
This worked for me when i ran this framework with the master IP of one
installed with DC/OS. But when I manually compiled Mesos (1.4/1.5) and started
agent/master on another machine I face this issue : "Implicitly declining
offers in ACCEPT call as the launch operation specified no tasks."
Configuration:
Master -
./bin/mesos-master.sh pas-lab-server6:5050 --work_dir=/home/sayantan/mesos_temp
--filter_gpu_resources=false
Agent (on the same node) -
./bin/mesos-agent.sh --master=pas-lab-server6:5050
--work_dir=/home/sayantan/mesos_temp
--isolation="filesystem/linux,cgroups/devices,gpu/nvidia"
I1010 14:48:44.106006 20372 sched.cpp:336] New master detected at
[email protected]:5050
I1010 14:48:44.107053 20372 sched.cpp:351] No credentials provided. Attempting
to register without authentication
I1010 14:48:44.113246 20365 sched.cpp:747] Framework registered with
00afc79a-cb51-4f9d-8c78-6000253ae396-0000
Registered!
.Starting task 0 on pas-lab-server6
Hello
Task 0 is in state 3
. . . . . .
The log on master says:
{code:java}
I1010 14:48:44.420087 19698 master.cpp:7766] Sending 1 offers to framework
00afc79a-cb51-4f9d-8c78-6000253ae396-0000 (Docker No Executor Framework (C++))
at [email protected]:45628
W1010 14:48:44.422901 19701 master.cpp:3911] Implicitly declining offers: [
00afc79a-cb51-4f9d-8c78-6000253ae396-O1 ] in ACCEPT call for framework
00afc79a-cb51-4f9d-8c78-6000253ae396-0000 as the launch operation specified no
tasks
I1010 14:48:44.423539 19701 master.cpp:9389] Removing offer
00afc79a-cb51-4f9d-8c78-6000253ae396-O1
I1010 14:48:44.423607 19701 master.cpp:4196] Processing ACCEPT call for offers:
[ 00afc79a-cb51-4f9d-8c78-6000253ae396-O1 ] on agent
00afc79a-cb51-4f9d-8c78-6000253ae396-S0 at slave(1)@127.0.1.1:5051
(pas-lab-server6) for framework 00afc79a-cb51-4f9d-8c78-6000253ae396-0000
(Docker No Executor Framework (C++)) at
[email protected]:45628
I1010 14:48:44.425109 19691 hierarchical.cpp:887] Updated allocation of
framework 00afc79a-cb51-4f9d-8c78-6000253ae396-0000 on agent
00afc79a-cb51-4f9d-8c78-6000253ae396-S0 from gpus(allocated: *):2;
cpus(allocated: *):40; mem(allocated: *):63421; disk(allocated: *):868675;
ports(allocated: *):[31000-32000] to gpus(allocated: *):2; cpus(allocated:
*):40; mem(allocated: *):63421; disk(allocated: *):868675; ports(allocated:
*):[31000-32000]
{code}
> Implicitly declining offers in ACCEPT call as the launch operation specified
> no tasks - for GPU based job
> ---------------------------------------------------------------------------------------------------------
>
> Key: MESOS-8065
> URL: https://issues.apache.org/jira/browse/MESOS-8065
> Project: Mesos
> Issue Type: Bug
> Components: docker, framework, gpu
> Affects Versions: 1.4.0
> Environment: Ubuntu 14.04
> Reporter: Ambuj Pandey
> Labels: docker, gpu, newbie
>
> I tried changing the "docker-no-executor-framework" to require GPU resource
> and use mesos containerizer with a docker image of nvidia/cuda.
> This worked for me when i ran this framework with the master IP of one
> installed with DC/OS. But when I manually compiled Mesos (1.4/1.5) and
> started agent/master on another machine I face this issue : "Implicitly
> declining offers in ACCEPT call as the launch operation specified no tasks."
> I've tested the nvidia/cuda image to work on that target machine using
> nvidia-docker.
> Configuration:
> Master -
> ./bin/mesos-master.sh pas-lab-server6:5050
> --work_dir=/home/sayantan/mesos_temp --filter_gpu_resources=false
> Agent (on the same node) -
> ./bin/mesos-agent.sh --master=pas-lab-server6:5050
> --work_dir=/home/sayantan/mesos_temp
> --isolation="filesystem/linux,cgroups/devices,gpu/nvidia"
> I1010 14:48:44.106006 20372 sched.cpp:336] New master detected at
> [email protected]:5050
> I1010 14:48:44.107053 20372 sched.cpp:351] No credentials provided.
> Attempting to register without authentication
> I1010 14:48:44.113246 20365 sched.cpp:747] Framework registered with
> 00afc79a-cb51-4f9d-8c78-6000253ae396-0000
> Registered!
> .Starting task 0 on pas-lab-server6
> Hello
> Task 0 is in state 3
> . . . . . .
> The log on master says:
> {code:java}
> I1010 14:48:44.420087 19698 master.cpp:7766] Sending 1 offers to framework
> 00afc79a-cb51-4f9d-8c78-6000253ae396-0000 (Docker No Executor Framework
> (C++)) at [email protected]:45628
> W1010 14:48:44.422901 19701 master.cpp:3911] Implicitly declining offers: [
> 00afc79a-cb51-4f9d-8c78-6000253ae396-O1 ] in ACCEPT call for framework
> 00afc79a-cb51-4f9d-8c78-6000253ae396-0000 as the launch operation specified
> no tasks
> I1010 14:48:44.423539 19701 master.cpp:9389] Removing offer
> 00afc79a-cb51-4f9d-8c78-6000253ae396-O1
> I1010 14:48:44.423607 19701 master.cpp:4196] Processing ACCEPT call for
> offers: [ 00afc79a-cb51-4f9d-8c78-6000253ae396-O1 ] on agent
> 00afc79a-cb51-4f9d-8c78-6000253ae396-S0 at slave(1)@127.0.1.1:5051
> (pas-lab-server6) for framework 00afc79a-cb51-4f9d-8c78-6000253ae396-0000
> (Docker No Executor Framework (C++)) at
> [email protected]:45628
> I1010 14:48:44.425109 19691 hierarchical.cpp:887] Updated allocation of
> framework 00afc79a-cb51-4f9d-8c78-6000253ae396-0000 on agent
> 00afc79a-cb51-4f9d-8c78-6000253ae396-S0 from gpus(allocated: *):2;
> cpus(allocated: *):40; mem(allocated: *):63421; disk(allocated: *):868675;
> ports(allocated: *):[31000-32000] to gpus(allocated: *):2; cpus(allocated:
> *):40; mem(allocated: *):63421; disk(allocated: *):868675; ports(allocated:
> *):[31000-32000]
> {code}
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
