[
https://issues.apache.org/jira/browse/YARN-9564?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16899283#comment-16899283
]
Eric Yang commented on YARN-9564:
---------------------------------
[~ebadger] Thank you for the patch.
# I think this is a user facing script, the source location for user facing
script resides in hadoop-yarn/bin. It maybe better to relocate the script to
this location. I am also concerned that there is no python user facing script
in Hadoop today. This script does not follow the Hadoop command line practice.
I am not sure if it is possible to create a shell wrapper on yarn cli to
invoke this script from libexec to make this script less stand out, but I let
others weight in on usability aspect of this script.
# If subprocess raised exception, process.kill() should be called to ensure
there is no orphan child process hanging around.
# The script accepts hadoop_prefix, but the hadoop shell commands are using
binary on $PATH variable. This might not work out of the box. Hadoop shell
script may help the bootstrap process by creating hadoop environment variables
to work with the python script to provide a better user experience for Hadoop
deployment that doesn't rely on PATH variable.
# Not sure how to work with working_dir parameter, it seems to fail if /tmp is
specified. Unless user give /tmp/test and test must not exist.
# This script requires password less sudo. This is usually a no-no from
Enterprise user. What can we do about these sudo commands? Maybe assume this
script require admin privileges?
# Missing Apache license header.
# How about having this writing in Java?
# get_local_manifest_from_path didn't close file descriptor.
# This utility depends on skopeo and python libraries. We may need a way to
manage python dependencies, if no plan on rewriting this in Java. At minimum
document the dependency package and version number required to run this script
as comments in the script.
> Create docker-to-squash tool for image conversion
> -------------------------------------------------
>
> Key: YARN-9564
> URL: https://issues.apache.org/jira/browse/YARN-9564
> Project: Hadoop YARN
> Issue Type: Sub-task
> Reporter: Eric Badger
> Assignee: Eric Badger
> Priority: Major
> Attachments: YARN-9564.001.patch
>
>
> The new runc runtime uses docker images that are converted into multiple
> squashfs images. Each layer of the docker image will get its own squashfs
> image. We need a tool to help automate the creation of these squashfs images
> when all we have is a docker image
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
