Reserve IP addresses for Slurm instances. #149
Conversation
wiktorn
force-pushed
the
reserve_ip_addresses
branch
from
c40beda to
6134ae8
Compare
| subnetwork = each.value.subnetwork | ||
| address_type = "INTERNAL" | ||
| region = var.region | ||
| address = each.value.network_ip |
There was a problem hiding this comment.
Isn't project should be set to handle SharedVPC?
I guess it should be a project from subnetwork.
Caveat: we never provide var.subnetwork_project, instead subnetwork is always a selfLink, that is sufficient for the providers.
In your case you would need to:
- Rename
var.subnetwork -> var.subnetwork_self_link - And either parse selfLink or create data source to extract project_id.
There was a problem hiding this comment.
The project doesn't have to (or even can't?) be in the SharedVPC. For sure what worked for me, creating them in service project. And I think it is better from management perspective (you see all the reserved IP addresses in service project).
But I'll do the renames and remove subnetwork_project and see how it works.
There was a problem hiding this comment.
Rename in _slurm_instance will require aligning the call sites of the module in cluster-toolkit.
Preliminary version of changes is available here
There was a problem hiding this comment.
Per discussion, I rolled back renaming subnetwork to subnetwork_self_link, but I dropped the subnetwork_project.
Should I mark it as DEPRECATED instead of silently dropping, as it is done in Cluster Toolkit?
Reserve IP addresses using google_compute_address for all Slurm instances so during the reconfiguration they maintain the same IP address, regardless if static IP addresses where provided or not.
wiktorn
force-pushed
the
reserve_ip_addresses
branch
from
6134ae8 to
449951d
Compare
|
@wiktorn , |
Reserve IP addresses using google_compute_address for all Slurm instances so during the reconfiguration they maintain the same IP address, regardless if static IP addresses where provided or not.