wiki instructions for Option 2 fix of (3.8.0 ‐ 3.9.3) ParallelCluster Build Image Failing during Installation of Minitar Ruby Gem Dependency aren't quite right.

[gwolski]

I tried following the option 2 instructions on the wiki page https://github.com/aws/aws-parallelcluster/wiki/(3.8.0-%E2%80%90-3.9.3)-ParallelCluster-Build-Image-Failing-during-Installation-of-Minitar-Ruby-Gem-Dependency but it failed. Here is my result and what I did that seems to work but needs validating by someone more familiar with this code section and then update wiki:

The code snippet for Option 2 shows:

## Replace below Line in InstallCinc Step
/opt/cinc/embedded/bin/gem install --local --no-document berkshelf:{{ BerkshelfVersion }}

## Replace with below line, effectively Pin the Minitar version to 0.12 
/opt/cinc/embedded/bin/gem install --no-document minitar:0.12
/opt/cinc/embedded/bin/gem install --local --no-document berkshelf:{{ BerkshelfVersion }}

However when I downloaded the source for parallelcluster 3.9.3, the original line that I'm replacing is missing the --local:

/opt/cinc/embedded/bin/gem install --no-document berkshelf:{{ BerkshelfVersion }}

Nonetheless, I replaced that one line with the exact code provided.

I then activated my pcluster 3.9.3 virtual env and ran the shown
pip install ./cli

My build-image failed.

I built a new virtual environment using the same pip install command after activating the new environment, still build failure.

The error message is of the form as shown on the wiki:

"Workflow Execution ID: 'wf-0988f3c0-6389-4138-bf1a-ceffc6b10876' failed with reason: Document arn:aws:imagebuilder:us-west-2:326469498578:component/parallelclusterimage-694714c0-99ae-11ef-8f75-060511802875/3.9.3/1 failed!"

Which suggested this error is still there - but I know I've got the right minitar now...

I know nothing about "gem install", but after talking with ChatGPT a bit, it seems adding the --local to the berkshelf install does not make sense. It wasn't there originally, we haven't downloaded berkshelf locally that I can figure out, so I tried removing the --local for the berkshelf install. ChatGPT tells me that berkshelf will use the minitar:0.12 we installed locally unless berkshelf specifies something specific that precludes it and since berkshelf isn't local, it will grab it from the proper remote repo.

So I changed the code to be:

              /opt/cinc/embedded/bin/gem install --no-document minitar:0.12
              /opt/cinc/embedded/bin/gem install --no-document berkshelf:{{ BerkshelfVersion }}

did the

pip install ./cli

and now my pcluster build-image actually created an ami and I can boot it standalone. Now to deploy it.

Was this posted code actually tested? Did I do this right?
As to moving forward with 3.10.x or 3.11.x - I'm having problems with both of those as well.
3.11.1 issue has been filed as #6529
and
3.10.1 issue might be an issue with aws-eda-slurm-cluster CloudFormation code, but I link it here for your review in case you have seen this as well: aws-samples/aws-eda-slurm-cluster#280

[himani2411]

Hi @gwolski

Thank you for catching this documentation mistake.
I have updated the https://github.com/aws/aws-parallelcluster/wiki/(3.8.0-%E2%80%90-3.9.3)-ParallelCluster-Build-Image-Failing-during-Installation-of-Minitar-Ruby-Gem-Dependency.

We did internally test the changes for these versions and they were successful. This was a documentation mistake made w.r.t code from 3.10.0 which is the version in which we added the --local flag for gem installation where we install some gems just before we install Minitar and Berkself.

[himani2411]

Will take a look at the other issues.

[gwolski]

Thank you for your confirmation and root causing how the error came to be. I have now been able to deploy it.