pcluster delete hangs when pcluster's IAM role has been modified by user

aws / aws-parallelcluster

AWS ParallelCluster is an AWS supported Open Source cluster management tool to deploy and manage HPC clusters in the AWS cloud.

https://github.com/aws/aws-parallelcluster

Apache License 2.0

828 stars 313 forks source link

pcluster delete hangs when pcluster's IAM role has been modified by user #1412

Closed aerogt3 closed 4 months ago

aerogt3 commented 4 years ago

Environment:

AWS ParallelCluster 2.4.1
OS: ubuntu 1604
Scheduler: slurm
Master instance type: c5n.2xlarge
Compute instance type: c5n.18xlarge

Bug description and how to reproduce: If you change the IAM role created by pcluster, pcluster delete hangs. Ideally it would still work, but at minimum would be nice to have error output.

Very easy to reproduce, just launch pcluster create, make a change to the IAM role, and run pcluster delete. In my case, I had granted read access to codecommit.

lukeseawalker commented 4 years ago

Hi @aerogt3, this is the expected behaviour of CloudFormation. You can verify that going through the "AWS console - CloudFormation", select the Cloudformation stack of the cluster and then "Stack Action - Detect drift". Right after click on "Stack Action - View drift results". The drift detection enables you to detect whether a stack's actual configuration has drifted from its template configuration.

We'll work to improve the error message for the delete command.

aerogt3 commented 4 years ago

Ok, understood. Thansk for the reply! In that case, as a feature suggestion, it would be nice to have a "--force" option for pcluster delete that removes the cluster despite any stack modification by the user.

enrico-usai commented 4 months ago

Marking as closed based on the fact that this matches an expected behaviour of CloudFormation that is designed to protect users against inadvertent errors. Users can "force" deletion from CloudFormation console, after checking the drifted resources.