[
https://issues.apache.org/jira/browse/YARN-7654?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16471073#comment-16471073
]
Eric Yang commented on YARN-7654:
---------------------------------
[~jlowe] I am struggling withe the following problems:
{quote}AbstractProviderService#buildContainerLaunchContext so the pieces needed
by DockerProviderService can be reused without requiring the launcher command
to be clobbered afterwards?{quote}
Launch command is override to bash -c 'launch-command' in
DockerLinuxContainerRuntime, and subsequently appended the log redirection. '2>
<LOG_DIR>/stderr.txt 1> <LOG_DIR>/stdout.txt', then replaced <LOG_DIR> with
actual container logging directory. The number of steps to go through the
preprocessing before writing to .cmd file complicates how to refactor the code
base without breaking things. This is the reason that setCommand was created
to flush out the override commands to ensure the command is not tempered
incorrectly during the hand off from DockerLinuxContainerRuntime to
DockerClient to container-executor. For safety reason, I keep setCommand to
ensure the command is not tempered by string substitutions, and not break YARN
v2 API.
{quote}The instance checking and downcasting in writeCommandToTempFile looks
pretty ugly. It would be cleaner to encapsulate this in the DockerCommand
abstraction. One example way to do this is to move the logic of writing a
docker command file into the DockerCommand abstract class. DockerRunCommand can
then override that method to call the parent method and then separately write
the env file. Worst case we can add a getEnv method to DockerCommand that
returns the collection of environment variables to write out for a command.
DockerCommand would return null or an empty collection while DockerRunCommand
can return its environment.{quote}
DockerCommand is a data structure class. It does not handle IO operation. If
we move IO operation to this class, it would not be clean data structure to
represent the docker command. I think it is more self explanatory that for
DockerRunCommand, we also write out the environment file. With changes in
YARN-8261, we are interested to ensure that directory is created, create the
cmd file, create the env file. For safety reason, I think we should not make
the styling changes for this area at this time because we are out of time to
throughly retest what have been tested in the previous patch set.
> Support ENTRY_POINT for docker container
> ----------------------------------------
>
> Key: YARN-7654
> URL: https://issues.apache.org/jira/browse/YARN-7654
> Project: Hadoop YARN
> Issue Type: Sub-task
> Components: yarn
> Affects Versions: 3.1.0
> Reporter: Eric Yang
> Assignee: Eric Yang
> Priority: Blocker
> Labels: Docker
> Attachments: YARN-7654.001.patch, YARN-7654.002.patch,
> YARN-7654.003.patch, YARN-7654.004.patch, YARN-7654.005.patch,
> YARN-7654.006.patch, YARN-7654.007.patch, YARN-7654.008.patch,
> YARN-7654.009.patch, YARN-7654.010.patch, YARN-7654.011.patch,
> YARN-7654.012.patch, YARN-7654.013.patch, YARN-7654.014.patch,
> YARN-7654.015.patch, YARN-7654.016.patch, YARN-7654.017.patch,
> YARN-7654.018.patch, YARN-7654.019.patch, YARN-7654.020.patch,
> YARN-7654.021.patch
>
>
> Docker image may have ENTRY_POINT predefined, but this is not supported in
> the current implementation. It would be nice if we can detect existence of
> {{launch_command}} and base on this variable launch docker container in
> different ways:
> h3. Launch command exists
> {code}
> docker run [image]:[version]
> docker exec [container_id] [launch_command]
> {code}
> h3. Use ENTRY_POINT
> {code}
> docker run [image]:[version]
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
