Not mounting FSx with ubuntu16.04

[JiaweiZhuang]

Environment:

• AWS ParallelCluster 2.2.1
• OS: ubuntu 16.04
• Scheduler: Slurm
• Master instance type: c5n.large
• Compute instance type: c5n.18xlarge

Bug description and how to reproduce:
With base_os = ubuntu1604, FSx was created but not mounted to the cluster. Does it currently require centos7?

For example, centos7 would install Lustre client during launch as shown in /var/log/cfn-init.log:

Recipe: aws-parallelcluster::_lustre_install
  * remote_file[/opt/parallelcluster/sources/kmod-lustre-client-2.10.6.x86_64.rpm] action create (skipped due to not_if)
  * remote_file[/opt/parallelcluster/sources/lustre-client-2.10.6.x86_64.rpm] action create (skipped due to not_if)
  * yum_package[lustre_kmod] action install (up to date)
  * yum_package[lustre_client] action install (up to date)

With ubuntu, it doesn't seem to do anything with Lustre:

• cfn-init.log (from ubuntu1604)

[JiaweiZhuang]

I just notice Support for FSx Lustre with Centos 7 in release notes. Doesn't seem to be mentioned in the docs though.

[demartinofra]

Hi @JiaweiZhuang

Yes FSx is only supported with Centos 7 in the current release.
I will update the docs accordingly and ideally we should also add a validation before cluster creation.

Thanks for reporting this!

[JiaweiZhuang]

Yes I can totally understand this because supporting other OS should not delay the current release. That's already a significant enhancement and thanks for the great work :)

[sean-smith]

Fixed in #904

We added validation to detect an unsupported instance type. Stayed tuned for more instance types soon!