terraform-provider-vsphere
terraform-provider-vsphere copied to clipboard
Add support for NVME controllers
Description
Hi there,
I've created a template using Packer with three attached NVME disks.
When I try to run Terraform for deploying an instance using this template, I get the message:
Error: there must be at least one disk specified
It seems that Terraform is not able to find the NVME disks attached (since the only opinions are pvscsi, buslogic, etc, but none NVME), even though I add an extra_config informing NVME is presenting:
disk_size_gb = var.vmRootDiskSize scsi_type = var.vmScsiType extra_config = { "nvme0.present" = "True" "nvme0.0.present" = "True"
PS: If we deploy a new instance using the Template manually, the instance boots up as expected.
References
As per the documentation, scsi_type NVME is not supported. https://registry.terraform.io/providers/hashicorp/vsphere/latest/docs/resources/virtual_machine#scsi_type
Community Note
Is the support for NVME on the roadmap (or maybe there's a workaround for now)?
Thanks for any suggestion or advice,
Hutger
I do not think it makes sense to extend scsi_type
to include NVME. it will need it's own nvme_
options like ide_
& sata_
.
Hi @apollo13 , would you propose an alternative approach for rolling out instances with NVME attached?
I don't think it is currently possibly with the terraform provider. If you do terraform import
of an existing machine with nvme disks you will see that those are not discovered at all. I fear there are plenty of changes needed :/
+1 for adding support for NVME controller (We would like to change the default from pvscsi to nvme in the future)
I am needing to recreate my template (made in Packer, also a HashiCorp project) to not use nvme because of this issue
When will this be supported?
Any update on this?
This is going to become a thing as NVME becomes the standard. Is support coming on this?
+1 this would be really useful
hi @tenthirtyam, any chance this can be re-prioritized? nvme is available to be built via packer but not deployed via terraform. im being asked to implement this. this ticket took over a year to get acknowledged.
thanks!
Hi @tenthirtyam, I'd like to 2nd @johnypony3 request. Our understanding of the nvme controller is that it improves performance when the backend disk is NVMe SSD. We are trying to build high performance Oracle and MS SQL database systems.
Acknowledged. This can be researched for an upcoming milestone based on the number of community reactions 👍.
If you have not added your own 👍, please do so on the originating description.
Ryan Johnson Staff II Solutions Architect | VMware, Inc.
+1 for adding support for NVME controller
+1 for adding support for NVME controller :thumbsup:
@patrikcze and @thewayofvikings - upvotes are only counted on the opening description, not comments.
Ryan Johnson Staff II Solutions Architect | VMware, Inc.
Hi, is there a possibility to plan this feature for 2.4 ? It's a big show stopper for us to use Terraform in prod because nvme is our standard. Would be amazing to have it. thanks
Acknowledged. This can be researched for an upcoming milestone based on the number of community reactions 👍.
If you have not added your own 👍, please do so on the originating description.
Ryan Johnson Staff II Solutions Architect | VMware, Inc.
This seems to be the most upvoted open issue in the repository at the moment, it would be great if it could be reconsidered for inclusion in a release milestone.
I'm still in favor of adding support for the NVMe controller, but I just wanted to point out that depending on what backend storage is being used, the use of this controller can actually lead to translation back to SCSI, thus negatively impacting performance. So check with your storage vendor before selecting this controller.
Any Update ??
Any Update ??
Please refer to the attached milestone for this enhancement request.
Is there really no update? I know the previous update says to look at the attached milestone, but I see that this was on the 2.3 milestone which has been completed. This is basically putting implementation of terraform on complete hold. It's very concerning to see the lack of meaningful development on this provider.
It was added to the backlog in 2022. Not really fair to say why wasn't it part of it 2.3 or even 2.4 when it's part of the backlog.
Remember, you are also allowed to contribute ;)
Yeah I get that contributions are welcome, but based on the backlog of pull requests, it doesn't seem like there's a lot of movement on this provider besides small bug fixes.
any news?
After a year, this is still the most upvoted open issue in the repository, and by a decent margin too. I must wonder why it's still stuck in the backlog after so long if issues are in fact prioritised by the number of 👍 reacts, as the issue template claims.
Sure, it would be nice if someone from the community would contribute this feature, but I'd still hope that a first-party, "official" provider would be better supported by HashiCorp, especially given that the README explicitly states that such providers are maintained by HashiCorp's Terraform team.