v3.0.1-rc1

[mleone87]

Ladies and gentlemen, the new release is almost ready!!

This RC1.

This version introduces breaking changes regarding the disks layout! Refer to documentation for further explanations

Please submit your feedbacks!

thanks to @Tinyblargon for the work on API side and on the provider side!

What's Changed

• Manage issues stale by @mleone87 in Manage issues stale #724
• docs: Remove deprecated preprovision section by @kencx in docs: Remove deprecated preprovision section #733
• fixes Destroy container fails with "500 CT not running" error #712 : destroy container fails "500 CT not running" by @btorresgil in fixes #712 : destroy container fails "500 CT not running" #728
• Add new vm_state attribute to manage running/stopped VMs by @jonglezb in Add new vm_state attribute to manage running/stopped VMs #725
• Smbios support for Qemu guest VM by @lsjostro in Smbios support for Qemu guest VM #736
• fix: Ignore disk resize if media is cdrom by @lsjostro in fix: Ignore disk resize if media is cdrom #737
• feat: update to go 1.19 by @mleone87 in feat: update to go 1.19 #740
• change PXE boot order and clean by @GannonTdW in change PXE boot order and clean #729
• Bump github.com/rs/zerolog from 1.29.0 to 1.29.1 by @dependabot in Bump github.com/rs/zerolog from 1.29.0 to 1.29.1 #746
• various stability & performance fixes by @mleone87 in various stability & performance fixes #749
• Update documentation with missing parameters for vm_qemu disk configuration by @AnotherIdea in Update documentation with missing parameters for vm_qemu disk configuration #756
• feat: cloud init disk resource by @lsjostro in feat: cloud init disk resource #748
• lxc: implement deletion of network interfaces by @GamerBene19 in lxc: implement deletion of network interfaces #754
• fix crash when disk.file is nil by @angelnu in fix crash when disk.file is nil #765
• Add disk parameters serial and wwn by @sergelogvinov in Add disk parameters serial and wwn #767
• docs: add smbios to qemu docs + example by @nepeat in docs: add smbios to qemu docs + example #776
• Bump github.com/kdomanski/iso9660 from 0.3.3 to 0.3.5 by @dependabot in Bump github.com/kdomanski/iso9660 from 0.3.3 to 0.3.5 #780
• Bump github.com/google/uuid from 1.3.0 to 1.3.1 by @dependabot in Bump github.com/google/uuid from 1.3.0 to 1.3.1 #834
• Bump github.com/kdomanski/iso9660 from 0.3.5 to 0.4.0 by @dependabot in Bump github.com/kdomanski/iso9660 from 0.3.5 to 0.4.0 #835
• Bump github.com/rs/zerolog from 1.29.1 to 1.31.0 by @dependabot in Bump github.com/rs/zerolog from 1.29.1 to 1.31.0 #845
• Bump github.com/hashicorp/terraform-plugin-sdk/v2 from 2.26.1 to 2.29.0 by @dependabot in Bump github.com/hashicorp/terraform-plugin-sdk/v2 from 2.26.1 to 2.29.0 #842
• feat: Add 'import-from' parameter to qemu by @jamesharr in feat: Add 'import-from' parameter to qemu #606
• Bump golang.org/x/net from 0.9.0 to 0.17.0 by @dependabot in Bump golang.org/x/net from 0.9.0 to 0.17.0 #873
• Bump google.golang.org/grpc from 1.57.0 to 1.57.1 by @dependabot in Bump google.golang.org/grpc from 1.57.0 to 1.57.1 #875
• Add initial efidisk support by @riverar in Add initial efidisk support #732
• don't reboot on update if automatic_reboot is disabled by @quantumsheep in don't reboot on update if automatic_reboot is disabled #759
• Added multiple target nodes instead of a single node. by @PascalKu in Added multiple target nodes instead of a single node. #766
• Fix Go module name (Invalid Go module #790) by @pandatix in Fix Go module name (#790) #791
• Adding the role SND.use required for lxc in Proxmox 8.0.3 by @hnrazevedo in Adding the role SND.use required for lxc in Proxmox 8.0.3 #809
• Allow the creation of bind mounts for lxc containers by @zinefer in Allow the creation of bind mounts for lxc containers #806
• added proxmox8 specific permission by @thigg in added proxmox8 specific permission #820
• Add ISO Storage resource by @h0lyalg0rithm in Add ISO Storage resource #822
• Make cores not default to 1 to allow for "unlimited" cores on containers by @devzeebo in Make cores not default to 1 to allow for "unlimited" cores on containers #829
• Fix lxc container size: Accept K by @Mice7R in Fix lxc container size: Accept K #836
• add validation to vm_qemu name field by @hedenface in add validation to vm_qemu name field #858
• Bugfix: reboot VM gracefully by @lucian-tx in Bugfix: reboot VM gracefully #861
• Feat: Gracefully shutdown VM on deletion. by @t3hpr1m3 in Feat: Gracefully shutdown VM on deletion. #872
• allow using dynamic list for ipconfig instead of hardcoding the index in the key by @spettinichi in allow using dynamic list for ipconfig instead of hardcoding the index in the key #866
• Bump golang.org/x/crypto from 0.16.0 to 0.17.0 by @dependabot in Bump golang.org/x/crypto from 0.16.0 to 0.17.0 #881
• Bump github.com/google/uuid from 1.4.0 to 1.5.0 by @dependabot in Bump github.com/google/uuid from 1.4.0 to 1.5.0 #880
• Bump github.com/hashicorp/terraform-plugin-sdk/v2 from 2.30.0 to 2.31.0 by @dependabot in Bump github.com/hashicorp/terraform-plugin-sdk/v2 from 2.30.0 to 2.31.0 #879
• Bugfix/require recent module by @MaartendeKruijf in Bugfix/require recent module #878
• resource_vm_qemu: allow . (dot) in name of virtual machine by @vilhelmprytz in resource_vm_qemu: allow . (dot) in name of virtual machine #886
• docs: New developer getting started guide by @hestiahacker in docs: New developer getting started guide #888
• Bump github.com/cloudflare/circl from 1.3.3 to 1.3.7 by @dependabot in Bump github.com/cloudflare/circl from 1.3.3 to 1.3.7 #894
• docs: added instructions on how to use a local library by @hestiahacker in docs: added instructions on how to use a local library #890
• feat: overhaul qemu disks implementation by @Tinyblargon in feat: overhaul qemu disks implementation #892
• chore: update modules by @mleone87 in chore: update modules #899
• Revert d02bb5d not properly tested by @mleone87 in Revert d02bb5d not properly tested #905
• chore: update dependencies by @mleone87 in chore: update dependencies #906

New Contributors

• @kencx made their first contribution in docs: Remove deprecated preprovision section #733
• @btorresgil made their first contribution in fixes #712 : destroy container fails "500 CT not running" #728
• @jonglezb made their first contribution in Add new vm_state attribute to manage running/stopped VMs #725
• @lsjostro made their first contribution in Smbios support for Qemu guest VM #736
• @AnotherIdea made their first contribution in Update documentation with missing parameters for vm_qemu disk configuration #756
• @GamerBene19 made their first contribution in lxc: implement deletion of network interfaces #754
• @angelnu made their first contribution in fix crash when disk.file is nil #765
• @sergelogvinov made their first contribution in Add disk parameters serial and wwn #767
• @nepeat made their first contribution in docs: add smbios to qemu docs + example #776
• @riverar made their first contribution in Add initial efidisk support #732
• @quantumsheep made their first contribution in don't reboot on update if automatic_reboot is disabled #759
• @PascalKu made their first contribution in Added multiple target nodes instead of a single node. #766
• @pandatix made their first contribution in Fix Go module name (#790) #791
• @hnrazevedo made their first contribution in Adding the role SND.use required for lxc in Proxmox 8.0.3 #809
• @zinefer made their first contribution in Allow the creation of bind mounts for lxc containers #806
• @thigg made their first contribution in added proxmox8 specific permission #820
• @h0lyalg0rithm made their first contribution in Add ISO Storage resource #822
• @devzeebo made their first contribution in Make cores not default to 1 to allow for "unlimited" cores on containers #829
• @Mice7R made their first contribution in Fix lxc container size: Accept K #836
• @hedenface made their first contribution in add validation to vm_qemu name field #858
• @lucian-tx made their first contribution in Bugfix: reboot VM gracefully #861
• @t3hpr1m3 made their first contribution in Feat: Gracefully shutdown VM on deletion. #872
• @spettinichi made their first contribution in allow using dynamic list for ipconfig instead of hardcoding the index in the key #866
• @MaartendeKruijf made their first contribution in Bugfix/require recent module #878
• @vilhelmprytz made their first contribution in resource_vm_qemu: allow . (dot) in name of virtual machine #886
• @hestiahacker made their first contribution in docs: New developer getting started guide #888

Full Changelog: v2.9.14...v3.0.1-rc1

---

This discussion was created from the release v3.0.1-rc1.

[hestiahacker]

I've done about 5 hours of testing and do think this release candidate should be promoted to an official release.

I have confirmed that there are some major issue with permissions in this release (namely #784 and #687), but they are also in the latest stable release, which is why I don't think it should hold up an official version 3.0.1.

[Jasmont]

In 2.9.14 there was an option for specifying volume for disks in proxmox_vm_qemu resource. Couldn't find it in this release. Are there any plans to add it?

[sajjadG]

That's the breaking change mentioned in the first post. Check the doc https://github.com/Telmate/terraform-provider-proxmox/blob/master/docs/resources/vm_qemu.md#disks-block

[electropolis]

I think unless there will be no resolution with disappearing of cloudinit drive that provides NOT a workaround but a stable solution that works with cicustom also, which I mention here #956 and attached my setup of cicustom to it by adjusting it to new requirements of this RC version (removed deprecated components), there should not be a stable release yet as it's for now still dropping cloudinit drive no matter what kind of workaround is added.

[hestiahacker]

It looks like this was fixed in the default branch, but hasn't been included in a 3.0.1 release candidate. So I believe if someone cuts a new release candidate from the master branch, we should be able to test it and hopefully confirm that the cicustom issue is no longer blocking the 3.0.1 release.

[spettinichi]

The hardcoded numbers in the discs block make this provider extremely difficult to use programmatically. We use a custom module to provision VMs with many server defaults, then in our main.tf we define each VM with only the configuration which varies from default. Dynamically setting disks when the number is hardcoded makes this much harder. We have run into the same issue with the netowrking with ipconfig0, etc. We effectively have to set one NIC in terraform and any additional are created manually after VM creation. I understand this change is required to some extent by changes in Proxmox, but using either a list or perhaps just repeated blocks would make this much more flexible.

Examples:

disks {
    scsi [
      {type=passthrough, etc},
      {type=disk, size=60gb, etc}
    ]
  }

disks {
  scsi {
    {type=passthrough, etc}
  }
  scsi {
    {type=disk, size=60gb, etc}
  }
}

[spettinichi]

@electropolis so changing it to be like this would work better for what you're trying to do with yaml? Cause it would also help me out for what I am doing

 disks {
    scsi {
      {
        name = scsi0
        cdrom {
          //<arguments omitted for brevity...>
        }
      }
      {
        name = scsi1
        disk {
          //<arguments omitted for brevity...>
        }
      }
      {
       name = scsi2
        passthrough {
          //<arguments omitted for brevity...>
        }
      }
      //<arguments omitted for brevity...>
    }
    //<arguments omitted for brevity...>
  }

[electropolis]

Exactly. It would be probably possible
With dynamic argument but I'm not sure. You can also try to put terragrunt on top and organise this in some kind of folder/file structure if you have any idea how terragrunt can be beneficial here.

[gbrackley]

@spettinichi I think this might be a balance of getting the normalisation/denormalisation right; and being able to round-trip that with the API.

The nested structure seems to be about:

• a bus type (scsi/ide/sata/virtio/...)
• a device type (cdrom/disk/passthrough)
• a device index/slot
• all the arguments that when omitted for brevity belie the residual complexity

Having a structured/specific slot has made the matching of devices for round-tripping with the API easier for the Terraform Provider, but made the resulting terraform less flexible and more structured for the user of the the provider.

I'm still really keen to see an example terraform module that allows all disk parameters to be passed through from a terraform script to the terraform provider. Expressing the 'disks' variable in the module, then passing the value provided by the user of that module to the provider seems to be thousands of lines of code.

[hestiahacker]

I'm fine with the schema as is and I'm also fine with the schema that @spettinichi proposed and @electropolis seconded. I have a slight preference for the latter even through it means updating all my terraform modules that deal with disks... again.

If someone is willing to implement the schema, I say we wait to release 3.0.1. If nobody is volunteering to write that code, then lets move ahead with what we already have in 3.0.1-rc1 so we can get all those bug fixes into an official release.

[spettinichi]

we ended up abandoning our module, i've mostly completed rewriting all of our terraform to remove it and be compliant with these changes. I took a look at the code and just don't have the time to learn the ins and outs of the proxmox api enough to re-write it and be confident enough to propose breaking it again. So i now agree with those who want to just cut a release.

I did find an unrelated bug which seems to already have an open ticket here but it hasn't been addressed #738 Given this breaking change I had to delete and re-import state, and this bug made that much more tedious than it should have been. I think it'd be nice to get it fixed for this since it is a breaking change, but don't think this should have to hold up the release necessarily.

[hestiahacker]

Is there anything we can do to help get an official v3.0.1 release, or is this just a matter of waiting until @mleone87 or @Tinyblargon has time to cut a release?

There are always going to be more tickets to handle, but it'd be nice to have all of these changes go out to people and handle additional feature requests, changes in functionality or bug fixes for a future release (perhaps v3.0.2, which maybe can quickly follow the v3.0.1 release).

The desire to get another release out, even if it's imperfect, was also expressed by @gbrackley over here

[greenship24]

I have done some testing. I can now reliably deploy a cluster of Fedora CoreOS VMs again. Using proxmox 8.1.3

Not getting any of the usual errors I did on 2.9.x now. Thanks for the hard work.

[kub3let]

I just did the upgrade and here are my 0.2$

• qemu-agent vm's with ipv6 only address currently break telmate see 3.0.1-rc1: Plan stuck on Refreshing state #978
• if using cicustom you must specify cloudinit_cdrom_storage now as well, this was not clear and not in the breaking changes, if cicustom is specified but cloudinit_cdrom_storage not it should fail with a error
• with the changes to bools in the disk configuration I think it would make sense to also specify agent and maybe other properties as bool instead of 0/1 int as well

[spettinichi]

i ran into this creating new VMs with 3.0.1-rc1, which I found is also in an issue here #727

There needs to be logs that cloudinit_cdrom_storage needs set if any cloudinit stuff is used but it is unset, or it needs to default to the first disk or something. Gotta dig to really figure out because the VM seems to start, it just never gets an IP set or anything. In addition to cicustom, it needs set if setting ipconfig or and login/ssh info.

[hestiahacker]

Just checking in again to see if this release can go out. There was a question about changing the schema, but that is no longer a blocker as people have hanged their terraform to match the API in this release.

It'd be good to release the improvements we already have done.

[kub3let]

There will be another release candidate #993

With all the feedback I assume a proper release will take more time.

Nothing prevents you from using the release candidate.