Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[FEA] Set a default Alluxio Master IP #5872

Closed
viadea opened this issue Jun 18, 2022 · 5 comments · Fixed by #5925
Closed

[FEA] Set a default Alluxio Master IP #5872

viadea opened this issue Jun 18, 2022 · 5 comments · Fixed by #5925
Assignees
Labels
P0 Must have for release task Work required that improves the product but is not user facing

Comments

@viadea
Copy link
Collaborator

viadea commented Jun 18, 2022

Currently we need to get the Alluxio Master IP and then use below Spark RAPIDS config to convert the path:

spark.rapids.alluxio.pathsToReplace="s3://foo->alluxio://RM_hostname:19998/foo,gs://bar->alluxio://RM_hostname:19998/bar"

I wish we can set a default Alluxio Master IP. (Say Master Node's IP) so that users do not need to manually find the Alluxio Master IP.
Eg, spark.rapids.alluxio.masterip=<Master Node's IP >.

@viadea viadea added feature request New feature or request ? - Needs Triage Need team to review and classify labels Jun 18, 2022
@sameerz sameerz added task Work required that improves the product but is not user facing and removed feature request New feature or request ? - Needs Triage Need team to review and classify labels Jun 21, 2022
@abellina
Copy link
Collaborator

abellina commented Jun 22, 2022

I am not entirely sure what that IP should be. Is there a guarantee that the master process is over by the Spark master (I assume that's what you are looking to do).

The config you mention here is verbose, especially since RM_hostname:19998 needs to be repeated for each mount. Could this be put into a different config so that the pathsToReplace one becomes: s3://foo->/foo,gs://bar->/bar or something like this?

@viadea
Copy link
Collaborator Author

viadea commented Jun 22, 2022

I am thinking if we can read the Alluxio conf dir, then we will know the alluxio master IP.
Then set it as default one spark.rapids.alluxio.masterip=<Master Node's IP >. unless users decide to change to a different master node IP.

Then as @abellina mentioned, the pathsToReplace can be much simpler.

@GaryShen2008 GaryShen2008 self-assigned this Jun 28, 2022
@GaryShen2008
Copy link
Collaborator

GaryShen2008 commented Jun 28, 2022

Since we use this config to support auto mount feature by calling "alluxio fs mount", I don't think we can support the case that Alluxio master is not the same node of spark driver node(mostly the spark master in Databricks case).
With this limitation, we can always get the alluxio master IP from alluxio's config file.
@viadea If you agree, I think it's no need to have this config. Instead, I'll have a new config like "spark.rapids.alluxio.automount.enabled" to enable the auto mount feature and get the master from alluxio's config.

@viadea
Copy link
Collaborator Author

viadea commented Jul 7, 2022

@GaryShen2008 I am thinking of some scenario:
If Alluxio master node is not one of the Spark nodes, then we could not find the alluxio's config. So how can we know where is the master? Do you think we can have a config spark.rapids.alluxio.masterip which can override the value?

@GaryShen2008
Copy link
Collaborator

@GaryShen2008 I am thinking of some scenario: If Alluxio master node is not one of the Spark nodes, then we could not find the alluxio's config. So how can we know where is the master? Do you think we can have a config spark.rapids.alluxio.masterip which can override the value?

In that case, even the current auto mount feature won't work since we call the command line to mount the folder.
If we want to support such case, we'll have to investigate if it's possible to use Alluxio JDK to mount remotely.

@tgravescs tgravescs added the P0 Must have for release label Jul 22, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
P0 Must have for release task Work required that improves the product but is not user facing
Projects
None yet
Development

Successfully merging a pull request may close this issue.

5 participants