new self-hosted runner is failing to start

[matanox]

Select Topic Area

Question

Copilot Feature Area

Copilot Agent Mode

Body

Is this a known issue that's new?

No files were found with the provided path: XXX/_work/_temp/runtime-logs/blocked.jsonl
XXX/_work/_temp/runtime-logs/blocked.md. No artifacts will be uploaded.

On the local machine running the runner, inside a docker image:

√ Connected to GitHub

Current runner version: '2.328.0'
2025-09-20 15:26:31Z: Listening for Jobs
2025-09-20 15:32:07Z: Running job: copilot
2025-09-20 15:32:41Z: Job copilot completed with result: Failed

The image is used for safety as I had the runner request sudo permissions before as much as I recall. The docker image is Ubuntu 24.04 based and is successfully used for verifying a complex bazel build.

I'm trying to reuse this well-tested image for running a self-hosted runner which then (if the runner is fixed to work) can facilitate copilot running those builds to verify its work, when copilot is dispatched at its "agent mode" on github.com.

Of course, my workflow file used by copilot is already configured to use my self-hosted runner, which as seen below it does successfully contact, but then it fails with the top of this message error shown in that agent detailed logs page, shown above.

[matanox]

Okay I figured after posting:

Running as non-root user and sudo not found . Add sudo configured for
password-less execution into the environment and try again. Exiting.
Error: Process completed with exit code 1.

But this requires more clicks in that web page shown above ― you have to click the top box with the wording "copilot" in it ― which is a little non-intuitive. Please improve the usability of that page for other future cases ...

The current workflow for zooming on agent errors is really prone to missing out.

[matanox]

Back to my original specific error:

• When I add the user to sudo, I get:

  $ ./run.sh 
  Must not run interactively with sudo
  

All-in-all it seems the runner won't agree to run as root, nor as a non-root user without sudo, nor as a non-root user granted sudo, as I tried all three options. If it's a specific permissions issue then the succinct error text does not tell more.

Perhaps someone can explain what I should do differently?

[matanox]

Digging in the _diag/ path under the work directory of the local runner:

[2025-09-20 15:32:35Z INFO ProcessInvokerWrapper] Starting process:
[2025-09-20 15:32:35Z INFO ProcessInvokerWrapper]   File name: '/foo/actions-runner/externals/node20/bin/node'
[2025-09-20 15:32:35Z INFO ProcessInvokerWrapper]   Arguments: '"/foo/actions-runner/_work/_actions/actions/upload-artifact/v4/dist/upload/index.js"'
[2025-09-20 15:32:35Z INFO ProcessInvokerWrapper]   Working directory: '/foo/actions-runner/_work/<repo-name>/<repo-name>'
[2025-09-20 15:32:35Z INFO ProcessInvokerWrapper]   Require exit code zero: 'False'
[2025-09-20 15:32:35Z INFO ProcessInvokerWrapper]   Encoding web name:  ; code page: ''
[2025-09-20 15:32:35Z INFO ProcessInvokerWrapper]   Force kill process on cancellation: 'False'
[2025-09-20 15:32:35Z INFO ProcessInvokerWrapper]   Redirected STDIN: 'False'
[2025-09-20 15:32:35Z INFO ProcessInvokerWrapper]   Persist current code page: 'False'
[2025-09-20 15:32:35Z INFO ProcessInvokerWrapper]   Keep redirected STDIN open: 'False'
[2025-09-20 15:32:35Z INFO ProcessInvokerWrapper]   High priority process: 'False'
[2025-09-20 15:32:35Z INFO ProcessInvokerWrapper] Updated oom_score_adj to 500 for PID: 386.
[2025-09-20 15:32:35Z INFO ProcessInvokerWrapper] Process started with process id 386, waiting for process exit.
[2025-09-20 15:32:35Z INFO JobServerQueue] Try to append 1 batches web console lines for record 'e7037eb0-dde6-4f21-900a-b5d51bc9029a', success rate: 1/1.
[2025-09-20 15:32:35Z INFO ProcessInvokerWrapper] STDOUT/STDERR stream read finished.
[2025-09-20 15:32:35Z INFO ProcessInvokerWrapper] STDOUT/STDERR stream read finished.
[2025-09-20 15:32:35Z INFO ProcessInvokerWrapper] Finished process 386 with exit code 0, and elapsed time 00:00:00.2201871.
[2025-09-20 15:32:35Z INFO CreateStepSummaryCommand] Step Summary file (/foo/actions-runner/_work/_temp/_runner_file_commands/step_summary_0edf2099-405c-4506-a858-77e4332d5130) is empty; skipping attachment upload
[2025-09-20 15:32:35Z INFO StepsRunner] Step result: 
[2025-09-20 15:32:35Z INFO ExecutionContext] Publish step telemetry for current step {
  "action": "actions/upload-artifact",
  "ref": "v4",
  "type": "node20",
  "stage": "Main",
  "stepId": "e7037eb0-dde6-4f21-900a-b5d51bc9029a",
  "stepContextName": "__actions_upload-artifact",
  "hasPreStep": false,
  "hasPostStep": false,
  "result": "succeeded",
  "errorMessages": [
    "No files were found with the provided path: /foo/actions-runner/_work/_temp/runtime-logs/blocked.jsonl\n/foo/actions-runner/_work/_temp/runtime-logs/blocked.md. No artifacts will be uploaded."
  ],
  "executionTimeInSeconds": 1,
  "startTime": "2025-09-20T15:32:35.6099691Z",
  "finishTime": "2025-09-20T15:32:35.8383326Z"
}.
[2025-09-20 15:32:35Z INFO StepsRunner] No need for updating job result with current step result 'Succeeded'.
[2025-09-20 15:32:35Z INFO StepsRunner] Current state: job state = 'Failed'
[2025-09-20 15:32:35Z INFO StepsRunner] Processing step: DisplayName='[Optional] Archive Details'
[2025-09-20 15:32:35Z INFO StepsRunner] Evaluating: (always() && ((((vars.COPILOT_AGENT_DEBUG == 'true') || ((vars.COPILOT_AGENT_DEBUG != 'true') && (vars.COPILOT_AGENT_DEBUG != 'false') && ('false' == 'true'))) && true) || false))
[2025-09-20 15:32:35Z INFO StepsRunner] Expanded: (true && ((((null == 'true') || ((null != 'true') && (null != 'false') && ('false' == 'true'))) && true) || false))
[2025-09-20 15:32:35Z INFO StepsRunner] Result: false
[2025-09-20 15:32:35Z INFO StepsRunner] Skipping step due to condition evaluation.
[2025-09-20 15:32:35Z INFO StepsRunner] No need for updating job result with current step result 'Skipped'.
[2025-09-20 15:32:35Z INFO StepsRunner] Current state: job state = 'Failed'
[2025-09-20 15:32:35Z INFO JobRunner] Finalize job.

I tried to pluck out the entire related block up here, which relates back to the start of this thread.

[matanox]

Should I supply more data from _work/ or _diag/?

The contents of .github/workflows/copilot-setup-steps.yml in the repo which copilot is unleashed in/from/on are:

name: Copilot Setup Steps

permissions:
  contents: read

jobs:
  copilot-setup-steps:
    runs-on: self-hosted
    steps:

      - name: Set up and enter a Python virtual environment
        run: |
          python3 -m venv .venv
          source .venv/bin/activate

[matanox]

Thanks Aqib, this is a lot of things to check, and I think I'll keep it for next times around.

Perhaps I should have installed the runner in the container as the non-root user to begin with and it is failing as it doesn't own all files/directories of the deflated runner files tree. 

For now I'll skip using AI as I'm helping improve runner ergonomics rather than gaining knowledge inside my actual target implementation domain. 

• I think the runner team could take only a little time to bake in some of the path of diagnostics which you mention into the runner's documentation, or even check and propagate errors more conducively for users up to the level of error text shown immediately by the runner
• Or any runner could point to where to look for details so that troubleshooting always starts at a better starting point when encountered.
• Every time I set up a self-hosted runner (this is not the first time) I ran into issues with it which took a non-linear trajectory to diagnose which the above suggestions would have likely avoided, or just made far more efficient.

I thank you for your comments, which should likely be pinned somewhere as a diagnostic how-to, but then, more automation and developer attention to details should avoid most of IMHO!

[matanox]

In one case few weeks back it was asking/using sudo, which implies using container isolation to me. However the current setup instruction page avoids detailing how to set up with containerized running ― that's a little too dead-ended to me ― as there should be a recipe in that page or anywhere at all for making it through to a secure container setup rather than only a cascade of error informational messages to bounce potential users of self-hosted runners around different ways towards successful operation ― even if they are a kind of dev-ops hardened :-) crowd of inidividuals.

Maybe I'm just missing something obvious or it's not really meant for github outsiders to use this option so much.