fix: wrong symlink path building for drive signing

[Gab-Menezes]

Fixes the symlink path for when signing by deviceId and vendorId.

Logs from the signing operation:

2026-04-07 04:09:53,994 - root - INFO - Found 0 drives with cluster_guid
2026-04-07 04:09:53,994 - root - INFO - Final excluded paths: []
2026-04-07 04:09:53,994 - root - INFO - Signing drives. Vendor ID: 144d, Device ID: a826
2026-04-07 04:09:53,995 - root - INFO - Running command: lspci -d 144d:a826 | sort | awk '{print $1}'
2026-04-07 04:09:54,012 - root - INFO - Command lspci -d 144d:a826 | sort | awk '{print $1}' finished with code 0
2026-04-07 04:09:54,012 - root - INFO - Command lspci -d 144d:a826 | sort | awk '{print $1}' stdout: 0000:1a:00.0
0000:1b:00.0
0000:2b:00.0
0000:2c:00.0

2026-04-07 04:09:54,012 - root - INFO - Running command: /weka-sign-drive sign -- /dev/disk/by-path/pci-0000:0000:1a:00.0-nvme-1 /dev/disk/by-path/pci-0000:0000:1b:00.0-nvme-1 /dev/disk/by-path/pci-0000:0000:2b:00.0-nvme-1 /dev/disk/by-path/pci-0000:0000:2c:00.0-nvme-1
2026-04-07 04:09:54,018 - root - INFO - Command /weka-sign-drive sign -- /dev/disk/by-path/pci-0000:0000:1a:00.0-nvme-1 /dev/disk/by-path/pci-0000:0000:1b:00.0-nvme-1 /dev/disk/by-path/pci-0000:0000:2b:00.0-nvme-1 /dev/disk/by-path/pci-0000:0000:2c:00.0-nvme-1 finished with code 1
2026-04-07 04:09:54,018 - root - INFO - Command /weka-sign-drive sign -- /dev/disk/by-path/pci-0000:0000:1a:00.0-nvme-1 /dev/disk/by-path/pci-0000:0000:1b:00.0-nvme-1 /dev/disk/by-path/pci-0000:0000:2b:00.0-nvme-1 /dev/disk/by-path/pci-0000:0000:2c:00.0-nvme-1 stdout: Processing 4 device(s): [/dev/disk/by-path/pci-0000:0000:1a:00.0-nvme-1 /dev/disk/by-path/pci-0000:0000:1b:00.0-nvme-1 /dev/disk/by-path/pci-0000:0000:2b:00.0-nvme-1 /dev/disk/by-path/pci-0000:0000:2c:00.0-nvme-1]

2026-04-07 04:09:54,018 - root - INFO - Command /weka-sign-drive sign -- /dev/disk/by-path/pci-0000:0000:1a:00.0-nvme-1 /dev/disk/by-path/pci-0000:0000:1b:00.0-nvme-1 /dev/disk/by-path/pci-0000:0000:2b:00.0-nvme-1 /dev/disk/by-path/pci-0000:0000:2c:00.0-nvme-1 stderr: Error processing /dev/disk/by-path/pci-0000:0000:2c:00.0-nvme-1: error checking device usage: failed to resolve device path: lstat /dev/disk/by-path/pci-0000:0000:2c:00.0-nvme-1: no such file or directory
Error processing /dev/disk/by-path/pci-0000:0000:1b:00.0-nvme-1: error checking device usage: failed to resolve device path: lstat /dev/disk/by-path/pci-0000:0000:1b:00.0-nvme-1: no such file or directory
Error processing /dev/disk/by-path/pci-0000:0000:1a:00.0-nvme-1: error checking device usage: failed to resolve device path: lstat /dev/disk/by-path/pci-0000:0000:1a:00.0-nvme-1: no such file or directory
Error processing /dev/disk/by-path/pci-0000:0000:2b:00.0-nvme-1: error checking device usage: failed to resolve device path: lstat /dev/disk/by-path/pci-0000:0000:2b:00.0-nvme-1: no such file or directory

The string 0000 is added wrongly leading the the wrong path, the command lspci -d {vendor_id}:{device_id} | sort | awk '{print $1}' already returns the 0000 prefix.

~$ lspci -d 144d:a826 | sort | awk '{print $1}'
0000:1a:00.0
0000:1b:00.0
0000:2b:00.0
0000:2c:00.0

Here is the WekaPolicy

apiVersion: weka.weka.io/v1alpha1
kind: WekaPolicy
metadata:
  name: sign-drives
  namespace: weka
spec:
  type: sign-drives
  payload:
    signDrivesPayload:
      type: device-identifiers
      pciDevices:
        vendorId: "144d"
        deviceId: "a826"
      nodeSelector:
        weka.io/supports-backends: "true"

[Copilot]

Pull request overview

Fixes incorrect /dev/disk/by-path symlink construction when signing drives discovered via lspci vendor/device IDs, preventing duplicate 0000: in the generated PCI path.

Changes:

• Update PCI by-path device string formatting to avoid prepending 0000: when lspci already includes the domain.
• Apply the same fix for both direct signing and proxy signing paths.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[graphite-app]

Graphite Automations

"Add anton/matt/sergey/kristina as reviwers on operator PRs" took an action on this PR • (04/07/26)

3 reviewers were added to this PR based on Anton Bykov's automation.

[Gab-Menezes]

hey @kristina-solovyova can we get this merged ? I would like very much for the next version of the operator to have this fixed.

Having to manually sign the drives because of this bug is a bit annoying.

[kristina-solovyova]

hey @kristina-solovyova can we get this merged ? I would like very much for the next version of the operator to have this fixed.

hay @Gab-Menezes! Thank you for the fix
Yes, I'm rebasing and merging it