[
https://issues.apache.org/jira/browse/MESOS-7361?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16243795#comment-16243795
]
Armand Grillet commented on MESOS-7361:
---------------------------------------
Different solutions exist:
# Log somewhere else than on the agent, this would require putting the UCR out
of the agent.
# Add an {{optional bool quiet = 4;}} defaulting to false to {{message
LaunchNestedContainerSession}} in {{agent.proto}} and checking it before
logging during the lifetime of the container. This would expose internal
"issues" to the API.
# Add a specific set of functions to the universal containerizer to handle the
lifecycle of the health checks that would do the same things as before but
quietly.
# Do not log to {{INFO}} messages coming from {{LaunchNestedContainerSession}},
we could have a parameter {{quiet}} implicitly set to true and having it set to
false when using {{LaunchNestedContainer}}.
> Command checks via agent pollute agent logs.
> --------------------------------------------
>
> Key: MESOS-7361
> URL: https://issues.apache.org/jira/browse/MESOS-7361
> Project: Mesos
> Issue Type: Improvement
> Components: agent
> Reporter: Alexander Rukletsov
> Assignee: Armand Grillet
> Labels: check, health-check, mesosphere
>
> Command checks via agent leverage debug container API of the agent to start
> checks. Each such invocation triggers a bunch of logs on the agent, because
> the API was not originally designed with periodic invocations in mind. We
> should find a way to avoid excessive logging on the agent.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
