[
https://issues.apache.org/jira/browse/YARN-9884?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16949497#comment-16949497
]
Eric Badger commented on YARN-9884:
-----------------------------------
bq. I am wondering if we need to combine the list to prevent future conflicts
of the same name?
I am fine with doing that. I refactored the error codes in a way that I thought
would be the least invasive to how docker currently handles error codes. But
having multiple error code lists is a little bit confusing because there could
be collisions with error values and you won't always know whether your error
came from a generic error code enum or a docker-specific one. Something I ran
into while testing was that the docker daemon will return codes in the 100
range. I had initially placed all of the docker error codes in the 100 range,
but then got an error 127 from the docker daemon and realized that that would
be incorrectly parsed as a docker error code. So even though the docker daemon
was passing back 127, the error message you would be given would be something
completely unrelated to the actual error.
So all in all, I am in favor of combining the lists. I'd be happy to include
that in this patch or put up a separate patch to do it.
> Make container-executor mount logic modular
> -------------------------------------------
>
> Key: YARN-9884
> URL: https://issues.apache.org/jira/browse/YARN-9884
> Project: Hadoop YARN
> Issue Type: Sub-task
> Reporter: Eric Badger
> Assignee: Eric Badger
> Priority: Major
> Attachments: YARN-9884.001.patch, YARN-9884.002.patch
>
>
> The current mount logic in the container-executor is interwined with docker.
> To avoid duplicating code between docker and runc, the code should be
> refactored so that both runtimes can use the same common code when possible.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
