[
https://issues.apache.org/jira/browse/MESOS-7918?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Philip Norman updated MESOS-7918:
---------------------------------
Description:
*Background*: This issue appeared while debugging dcos-metrics. A DC/OS cluster
running the HDFS framework was missing application metrics. The HDFS tasks were
emitting metrics. The metrics process was aggregating them. But since no entry
was available for them in the /containers endpoint, they were not discoverable.
*Specific bug*: The agent in question was running two executors and two tasks -
four containers in total. /containers only showed two containers. It only
showed the executor containers, not their nested task containers. Note the
disparity between the output of
[/state](https://gist.github.com/philipnrmn/4b68f72f7b2a50c885c16c8346582d35)
and
[/containers](https://gist.github.com/philipnrmn/955fa5049ac706ca252ca6fd9e549188).
*Suggested fix*: The containers endpoint should display a flattened list of all
the containers running on the agent.
*To reproduce this issue*: Start a framework which uses the default executor.
Check the /containers output of the agent on which it is running.
was:
h3. Background
This issue while debugging dcos-metrics. A DC/OS cluster running the HDFS
framework was missing application metrics. The HDFS tasks were emitting
metrics. The metrics process was aggregating them. But since no entry was
available for them in the /containers endpoint, they were not discoverable.
h3. What is wrong
The agent in question was running two executors and two tasks - four containers
in total. /containers only showed two containers. It only showed the executor
containers, not their nested task containers. Note the disparity between the
output of
[/state](https://gist.github.com/philipnrmn/4b68f72f7b2a50c885c16c8346582d35)
and
[/containers](https://gist.github.com/philipnrmn/955fa5049ac706ca252ca6fd9e549188).
h3. Suggested fix
The containers endpoint should display a flattened list of all the containers
running on the agent.
h3. To reproduce this issue
Start a framework which uses the default executor. Check the /containers output
of the agent on which it is running.
> The agent /containers endpoint does not return nested containers
> ----------------------------------------------------------------
>
> Key: MESOS-7918
> URL: https://issues.apache.org/jira/browse/MESOS-7918
> Project: Mesos
> Issue Type: Bug
> Components: HTTP API
> Environment: DC/OS
> Reporter: Philip Norman
>
> *Background*: This issue appeared while debugging dcos-metrics. A DC/OS
> cluster running the HDFS framework was missing application metrics. The HDFS
> tasks were emitting metrics. The metrics process was aggregating them. But
> since no entry was available for them in the /containers endpoint, they were
> not discoverable.
> *Specific bug*: The agent in question was running two executors and two tasks
> - four containers in total. /containers only showed two containers. It only
> showed the executor containers, not their nested task containers. Note the
> disparity between the output of
> [/state](https://gist.github.com/philipnrmn/4b68f72f7b2a50c885c16c8346582d35)
> and
> [/containers](https://gist.github.com/philipnrmn/955fa5049ac706ca252ca6fd9e549188).
> *Suggested fix*: The containers endpoint should display a flattened list of
> all the containers running on the agent.
> *To reproduce this issue*: Start a framework which uses the default executor.
> Check the /containers output of the agent on which it is running.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
