Unable to create cluster after upgrading to 2.2.1 - ERROR: Unable to check AWS Account limits

[tennex-jack]

Environment:

• AWS ParallelCluster version: aws-parallelcluster-2.2.1
• OS: alinux/MacOS

Bug description and how to reproduce:
pcluster create test
pcluster configure

After upgrading to 2.2.1 I am no longer able to create clusters, or even re-run the configure utility without encountering this error:

ERROR: Unable to check AWS Account limits. Please double check your cluster configuration.
No default VPC for this user

Additional context:
~/.parallelcluster/pcluster-cli.log :

2019-03-04 22:22:36,136 - DEBUG - pcluster.pcluster - Namespace(command='configure', config_file=None, func=<function configure at 0x7efe3b8f7938>)
2019-03-04 22:23:06,754 - DEBUG - pcluster.pcluster - pcluster CLI starting
2019-03-04 22:23:06,759 - DEBUG - pcluster.pcluster - Namespace(command='configure', config_file=None, func=<function configure at 0x7fad66286938>)
2019-03-04 22:23:48,901 - DEBUG - pcluster.pcluster - pcluster CLI starting
2019-03-04 22:23:48,907 - DEBUG - pcluster.pcluster - Namespace(command='configure', config_file=None, func=<function configure at 0x7fed693c5938>)

~/.parallelcluster/config:

aws_region_name = us-east-1

[cluster default]
key_name = <my_key>
vpc_settings = vpc

[vpc vpc]
vpc_id = vpc-d444ebaf
master_subnet_id = subnet-3e947c74
ssh_from = <my_ip>

[global]
cluster_template = default
update_check = true
sanity_check = true

[aliases]
ssh = ssh {CFN_USER}@{MASTER_IP} {ARGS}

AWS CLI:

aws --version
aws-cli/1.16.117 Python/3.6.2 Darwin/18.2.0 botocore/1.12.107

What I've tried:

• Cluster creation and general management worked fine on the previous version (2.0.0), this only began happening after the upgrade.
• Confirmed this both on my local machine (MacOS) and with an Amazon Linux 2 instance in AWS.
• I do not have a default VPC, but I also did not have one before the upgrade.
• I have tried this with IAM API Keys and an assumed role, both of which have AdministratorAccess

I know that with the new release, there was logic added to check AWS Account Limits. Could that be related? I hope I am doing something stupid and this is a quick fix! Please let me know if there's any further context I can provide.

Thanks,

Jack

[sean-smith]

Hi,

Please set sanity_check = false as a workaround so you can launch clusters. This is part of the logic we added to validate that you have the right limits.

[sean-smith]

Ok, the problem is coming from #L403

We do an ec2 dry-instance run to validate the account limits. It tries to use the default VPC as there's no VPC specified in the call. That's failing and we catch it and return that error.

For now the workaround is using sanity_check = false, until we fix it.

[enrico-usai]

Fixed by #911,
@privojack thank you for the feedback!

[hellmersjl3ds]

I'm trying to configure ParallelCluster for the first time, and I try to modify the config file to not do the sanity check, but it keeps setting it back. Is there a way to configure it on initial configuration so that it doesn't do the sanity check?

[enrico-usai]

Hi @hellmersjl3ds I created a new issue #913 for your last question since it is not strictly related to the AWS account limits check error.