Give a short summary of changes on ghpc deploy/destroy

[yaroslavvb]

Right now ghpc output is too verbose. Destroying a cluster like this yaml prints >40 pages of changes (my tmux ran out of buffer at 40 pages).

Ideally there's be a way to see what exactly is being destroyed that fits on a couple of screens

[nick-stroud]

Thank you for your bug report. I have investigated the issue and determined that it is not a bug, but rather a feature request.

I am moving this to the ideas-and-feature-request discussion where feature requests are collected.

One possible workaround would be to use a grep on terraform plan to summarize from a high level what is changing.

./ghpc create <blueprint> -w
terraform -chdir=<deployment>/<group> init
terraform -chdir=<deployment>/<group> plan | grep '# module'
# or alternatively for destroying
terraform -chdir=<deployment>/<group> plan -destroy | grep '# module'

This will reduce the output to 1 line for each resource changed. Example output for a random change I happen to be executing:

  # module.batch-job-base.local_file.job_template will be created
  # module.batch-job-batch.local_file.job_template will be created
  # module.batch-job-hpc.local_file.job_template will be created
  # module.batch-job-ubuntu.local_file.job_template will be created
  # module.batch-login.google_compute_instance_from_template.batch_login will be destroyed
  # module.wait2.null_resource.wait_for_startup will be destroyed
  # module.batch-login.module.login_startup_script.google_storage_bucket.configs_bucket[0] will be destroyed
  # module.batch-login.module.login_startup_script.google_storage_bucket_object.scripts["0-install_ddn_lustre_client.sh"] will be destroyed
  # module.batch-login.module.login_startup_script.google_storage_bucket_object.scripts["0-mount-ddn-lustre.sh"] will be destroyed
  # module.batch-login.module.login_startup_script.google_storage_bucket_object.scripts["1-install_ddn_lustre_client.sh"] will be destroyed
  # module.batch-login.module.login_startup_script.google_storage_bucket_object.scripts["1-mount-ddn-lustre.sh"] will be destroyed
  # module.batch-login.module.login_startup_script.google_storage_bucket_object.scripts["README.md"] will be destroyed
  # module.batch-login.module.login_startup_script.google_storage_bucket_object.scripts["cloud-batch-cent-base.json"] will be destroyed
  # module.batch-login.module.login_startup_script.google_storage_bucket_object.scripts["cloud-batch-cent-hpc.json"] will be destroyed
  # module.batch-login.module.login_startup_script.google_storage_bucket_object.scripts["passed_startup_script.sh"] will be destroyed
  # module.batch-login.module.login_startup_script.random_id.resource_name_suffix will be destroyed

[issacg]

I like the availability of the full output, but agree that a terse option would be good. Perhaps an addition option to d,a,s,c to show a terse summary for practitioners who use hpc-toolkit but don't actually know terraform (which I thought was part of the point of the toolkit? :))

However, for the long outputs when I touch something sensitive, I want to diff the whole thing. My workaround is to manually run terraform plan with less or another pager (and then enjoy the colors which also help me spot potential problems (such as create/destroy instead of modify)