LXD device gets corrupted when ssh connection is closed
Metadata
Current evaluation
Snapcraft LXD build instance fails to restart after SSH disconnect/VM reboot because the craft-state disk device points at a transient /run/user/<uid>/<pid> path that no longer exists. Unlabelled, zero comments, 245 days old.
Suggested action: needs triage
Reason: No labels, no assignee, and zero comments from any maintainer in 245 days, so this has never been triaged. The root cause is identifiable from the log — craft-application's state service (craft_application/services/state.py) places its state dir under $XDG_RUNTIME_DIR/<pid>, which is wiped on session end/reboot, leaving the LXD disk-/tmp/craft-state device with a missing source path. It is a real, reproducible bug (not a support question) but needs maintainer assessment of where the fix belongs (craft-application state dir location vs craft-providers device handling).
Impact:
50
Quick Win:
22.5
Staleness:
75
Complexity:
55
Confidence:
65
Support Request:
10
Issue body
### Check existing issues
- [x] I've verified that this bug isn't described by any existing issues.
### Bug description
I'm using a gcloud compute vm to build snaps. Whenever I am closing the ssh connection, or if the gcloud instance shuts down, and I restart it and continue the snapcraft build, LXD fails to find the old device.
### Steps to reproduce
1. Connect to a gcloud compute instance
2. Start Building the snap
3. Stop the build
4. Close the ssh connection and reconnect
5. See this error
### Environment
GCloud Compute instance
Snapcraft: version 8.13.2 revision 16570
LXD: version 5.21.4-9eb1368 revision 36971
### snapcraft.yaml
```yaml
```
### Log output
```shell
Starting instance
2025-12-29 06:54:28.292 Executing on host: lxc --project snapcraft info local:snapcraft-qt-full-amd64-524348
2025-12-29 06:54:28.408 Executing on host: lxc --project snapcraft start local:snapcraft-qt-full-amd64-524348
2025-12-29 06:54:28.540 Failed to start 'snapcraft-qt-full-amd64-524348'.
2025-12-29 06:54:28.540 Detailed information:
* Command that failed: 'lxc --project snapcraft start local:snapcraft-qt-full-amd64-524348'
* Command exit code: 1
* Command standard error output: b'Error: Failed start validation for device "disk-/tmp/craft-state": Missing source path "/run/user/400433789/356321" for disk "disk-/tmp/craft-state"\nTry `lxc info --show-log local:snapcraft-qt-full-amd64-524348` for more info\n'
2025-12-29 06:54:28.555 Traceback (most recent call last):
2025-12-29 06:54:28.555 File "/snap/snapcraft/16570/lib/python3.12/site-packages/craft_providers/lxd/lxc.py", line 1187, in start
2025-12-29 06:54:28.556 self._run_lxc(command, capture_output=True, project=project)
2025-12-29 06:54:28.556 File "/snap/snapcraft/16570/lib/python3.12/site-packages/craft_providers/lxd/lxc.py", line 206, in _run_lxc
2025-12-29 06:54:28.556 return subprocess.run(
2025-12-29 06:54:28.556 ^^^^^^^^^^^^^^^
2025-12-29 06:54:28.556 File "/snap/snapcraft/current/usr/lib/python3.12/subprocess.py", line 571, in run
2025-12-29 06:54:28.556 raise CalledProcessError(retcode, process.args,
2025-12-29 06:54:28.556 subprocess.CalledProcessError: Command '['lxc', '--project', 'snapcraft', 'start', 'local:snapcraft-qt-full-amd64-524348']' returned non-zero exit status 1.
2025-12-29 06:54:28.556
2025-12-29 06:54:28.556 The above exception was the direct cause of the following exception:
2025-12-29 06:54:28.557 Traceback (most recent call last):
2025-12-29 06:54:28.557 File "/snap/snapcraft/16570/lib/python3.12/site-packages/craft_application/application.py", line 663, in run
2025-12-29 06:54:28.557 return_code = self._run_inner()
2025-12-29 06:54:28.557 ^^^^^^^^^^^^^^^^^
2025-12-29 06:54:28.557 File "/snap/snapcraft/16570/lib/python3.12/site-packages/snapcraft/application.py", line 183, in _run_inner
2025-12-29 06:54:28.557 return_code = super()._run_inner()
2025-12-29 06:54:28.557 ^^^^^^^^^^^^^^^^^^^^
2025-12-29 06:54:28.557 File "/snap/snapcraft/16570/lib/python3.12/site-packages/craft_application/application.py", line 640, in _run_inner
2025-12-29 06:54:28.557 return_code = dispatcher.run() or os.EX_OK
2025-12-29 06:54:28.557 ^^^^^^^^^^^^^^^^
2025-12-29 06:54:28.557 File "/snap/snapcraft/16570/lib/python3.12/site-packages/craft_cli/dispatcher.py", line 564, in run
2025-12-29 06:54:28.558 return self._loaded_command.run(self._parsed_command_args)
2025-12-29 06:54:28.558 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-12-29 06:54:28.558 File "/snap/snapcraft/16570/lib/python3.12/site-packages/craft_application/commands/base.py", line 222, in run
2025-12-29 06:54:28.558 result = self._run(parsed_args, **kwargs) or result
2025-12-29 06:54:28.558 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-12-29 06:54:28.558 File "/snap/snapcraft/16570/lib/python3.12/site-packages/snapcraft/commands/lifecycle.py", line 74, in _run
2025-12-29 06:54:28.558 super()._run(parsed_args)
2025-12-29 06:54:28.558 File "/snap/snapcraft/16570/lib/python3.12/site-packages/craft_application/commands/lifecycle.py", line 619, in _run
2025-12-29 06:54:28.558 return super()._run(parsed_args=parsed_args, step_name=step_name)
2025-12-29 06:54:28.558 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-12-29 06:54:28.559 File "/snap/snapcraft/16570/lib/python3.12/site-packages/craft_application/commands/lifecycle.py", line 232, in _run
2025-12-29 06:54:28.559 self._run_manager_for_build_plan(fetch_service_policy)
2025-12-29 06:54:28.559 File "/snap/snapcraft/16570/lib/python3.12/site-packages/craft_application/commands/lifecycle.py", line 97, in _run_manager_for_build_plan
2025-12-29 06:54:28.559 provider.run_managed(build, bool(fetch_service_policy))
2025-12-29 06:54:28.559 File "/snap/snapcraft/16570/lib/python3.12/site-packages/craft_application/services/provider.py", line 457, in run_managed
2025-12-29 06:54:28.559 with self.instance(
2025-12-29 06:54:28.559 File "/snap/snapcraft/current/usr/lib/python3.12/contextlib.py", line 137, in __enter__
2025-12-29 06:54:28.559 return next(self.gen)
2025-12-29 06:54:28.559 ^^^^^^^^^^^^^^
2025-12-29 06:54:28.559 File "/snap/snapcraft/16570/lib/python3.12/site-packages/craft_application/services/provider.py", line 185, in instance
2025-12-29 06:54:28.559 with provider.launched_environment(
2025-12-29 06:54:28.559 File "/snap/snapcraft/current/usr/lib/python3.12/contextlib.py", line 137, in __enter__
2025-12-29 06:54:28.560 return next(self.gen)
2025-12-29 06:54:28.560 ^^^^^^^^^^^^^^
2025-12-29 06:54:28.560 File "/snap/snapcraft/16570/lib/python3.12/site-packages/craft_providers/lxd/lxd_provider.py", line 168, in launched_environment
2025-12-29 06:54:28.560 instance = launch(
2025-12-29 06:54:28.560 ^^^^^^^
2025-12-29 06:54:28.560 File "/snap/snapcraft/16570/lib/python3.12/site-packages/craft_providers/lxd/launcher.py", line 780, in launch
2025-12-29 06:54:28.560 if instance.exists() and _launch_existing_instance(
2025-12-29 06:54:28.560 ^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-12-29 06:54:28.560 File "/snap/snapcraft/16570/lib/python3.12/site-packages/craft_providers/lxd/launcher.py", line 429, in _launch_existing_instance
2025-12-29 06:54:28.560 instance.start()
2025-12-29 06:54:28.560 File "/snap/snapcraft/16570/lib/python3.12/site-packages/craft_providers/lxd/lxd_instance.py", line 568, in start
2025-12-29 06:54:28.560 self.lxc.start(
2025-12-29 06:54:28.561 File "/snap/snapcraft/16570/lib/python3.12/site-packages/craft_providers/lxd/lxc.py", line 1189, in start
2025-12-29 06:54:28.561 raise LXDError(
2025-12-29 06:54:28.561 craft_providers.lxd.errors.LXDError: Failed to start 'snapcraft-qt-full-amd64-524348'.
2025-12-29 06:54:28.561 * Command that failed: 'lxc --project snapcraft start local:snapcraft-qt-full-amd64-524348'
2025-12-29 06:54:28.561 * Command exit code: 1
2025-12-29 06:54:28.561 * Command standard error output: b'Error: Failed start validation for device "disk-/tmp/craft-state": Missing source path "/run/user/400433789/356321" for disk "disk-/tmp/craft-state"\nTry `lxc info --show-log local:snapcraft-qt-full-amd64-524348` for more info\n'
```
Evaluation history
| Date | Model | Scores | Action | Summary |
|---|---|---|---|---|
| qwen/qwen3.8-27b |
Impact:
50
Quick Win:
22.5
Staleness:
75
Complexity:
55
Confidence:
65
Support Request:
10
|
needs triage | Snapcraft LXD build instance fails to restart after SSH disconnect/VM reboot because the craft-state disk device points at a transient /run/user/<uid>/<pid> path that no longer exists. Unlabelled, zero comments, 245 days old. | |
| qwen/qwen3.6-35b-a3b |
Staleness:
90
Complexity:
60
Confidence:
85
Support Request:
10
|
close stale | LXD instance fails to start after SSH disconnect due to missing transient user runtime path. No maintainer activity or comments for 205 days. | |
| qwen3.6-35b-a3b-mtp-q6 |
Staleness:
75
Complexity:
40
Confidence:
85
Support Request:
10
|
needs triage | LXD device fails to start after SSH disconnect due to missing ephemeral source path. Untriaged, 191 days old, no maintainer response. | |
| qwen3.6-35b-a3b-mtp-q6 |
Staleness:
80
Complexity:
45
Confidence:
85
Support Request:
10
|
needs triage | LXD disk device fails to start after SSH disconnect or VM restart due to a missing temporary source path. Open for 169 days with zero comments or labels; needs triage and assignment. |
Update history
No update history recorded yet.
Related work
-
Related To:
craft-application#786
(confidence 60%)
feat: add a state service — introduced the state dir that is bind-mounted into the LXD instance as disk-/tmp/craft-state and sourced from XDG_RUNTIME_DIR, the path that goes missing.
-
Related To:
craft-application#838
(confidence 55%)
fix(StateService): add g+rwx to the state dir — shows the state dir location/permissions are an active area of change in the same service implicated here.
Related issues
No related issues found above the similarity threshold.