Error initializing block storage client with Openstack 2026.1

[elelaysh]

Overview of the Issue

With the 2026.1 (Gazpacho) release, cinder dropped the volumev3 endpoint.
It needs to be fetched from the endpoint list instead of hardcoded.

The error is Error initializing block storage client: No suitable endpoint could be found in the service catalog.

Reproduction Steps

After the upgrade of a cloud to 2026.1, packer build will fail when it previously worked:

# PACKER_LOG=1 /usr/bin/packer build   openstack.pkr.hcl
(...)
openstack.openhpc: output will be in this color.

2026/08/25 15:40:35 Running builder: openstack
2026/08/25 15:40:35 [INFO] (telemetry) Starting builder openstack.openhpc
==> openstack.openhpc: Loading flavor: general.v1.small
2026/08/25 15:40:35 packer-plugin-openstack_v1.1.4_x5.0_linux_amd64 plugin: 2026/08/25 15:40:35 [INFO] Loading flavor by ID: general.v1.small
2026/08/25 15:40:35 packer-plugin-openstack_v1.1.4_x5.0_linux_amd64 plugin: 2026/08/25 15:40:35 [ERROR] Failed to find flavor by ID: Resource not found
2026/08/25 15:40:35 packer-plugin-openstack_v1.1.4_x5.0_linux_amd64 plugin: 2026/08/25 15:40:35 [INFO] Loading flavor by name: general.v1.small
==> openstack.openhpc: Verified flavor. ID: d7d3ab48-3948-4140-a690-60626ed3cabe
==> openstack.openhpc: Using existing SSH private key
==> openstack.openhpc: Using existing SSH private key
==> openstack.openhpc: Error initializing block storage client: No suitable endpoint could be found in the service catalog.
==> openstack.openhpc: Step "StepCreateVolume" failed
==> openstack.openhpc: [c] Clean up and exit, [a] abort without cleanup, or [r] retry step (build may fail even if retry succeeds)?

Plugin and Packer version

Packer v1.16.0
packer-plugin-openstack v1.1.4

Simplified Packer Buildfile

packer {
  required_plugins {
    openstack = {
      version = ">= 1.0.0"
      source  = "github.com/hashicorp/openstack"
    }
  }
}
source "openstack" "openhpc" {

  flavor = "en1.small"
  use_blockstorage_volume = true
  volume_type = "ssd"
  volume_size = 20
  volume_name = "my-image-xxx"
  
  # Input image:
  source_image = "xxxx"
  source_image_name = "rl-98"
  
  # Output image:
  image_disk_format = "raw"
}

build {

  source "source.openstack.openhpc" {
    image_name = "my-image-xxx"
  }

}

Log Fragments and crash.log files

Openstack endpoints on 2025.1

$ openstack endpoint list | grep cinder
| uuid1 | RegionOne | cinderv3         | volumev3                | True    | public    | https://my.cloud.local/v3/%(tenant_id)s          |
| uuid2 | RegionOne | cinder           | block-storage           | True    | public    | https://my.cloud.local/v3                        |
| uuid3 | RegionOne | cinderv3         | volumev3                | True    | internal  | https://my.cloud.private.ip:8776/v3/%(tenant_id)s                             |
| uuid4 | RegionOne | cinder           | block-storage           | True    | internal  | https://my.cloud.private.ip:8776/v3

Openstack endpoints on 2026.1

| uuid5 | RegionOne | cinder           | block-storage           | True    | public    | https://my.cloud.private.ip:8776/v3                           |
| uuid6 | RegionOne | cinder           | block-storage           | True    | internal  | https://my.cloud.private.ip:8776/v3

==> volumev3 endpoints have been removed, must switch to block-storage

[elelaysh]

Solution is to upgrade to gophercloud/v2 and gophercloud/utils/v2.
This is mostly mechanical changes, with the need to pass a context on all gophercloud api calls.