자체 호스트형 실행기 모니터링 및 문제 해결

Using repository-level self-hosted runners

You may not be able to create a self-hosted runner for an organization-owned repository.

Enterprise owners and organization owners can choose which repositories are allowed to create repository-level self-hosted runners. Users with the “Manage organization runners and runner groups” permission can only choose which repositories are allowed to create repository-level self-hosted runners for repositories in your organization.

For more information about custom organization roles, see "About custom organization roles."

For more information, see "Enforcing policies for GitHub Actions in your enterprise" and "Disabling or limiting GitHub Actions for your organization."

Checking the status of a self-hosted runner

A self-hosted runner can be located in either your repository, organization, or enterprise account settings on GitHub. To manage a self-hosted runner, you must have the following permissions, depending on where the self-hosted runner was added:

User repository: You must be the repository owner.
Organization: You must be an organization owner.
Organization repository: You must be an organization owner, or have admin access to the repository.
Enterprise account: You must be an enterprise owner.

In your organization or repository, navigate to the main page and click Settings.
In the left sidebar, click Actions, then click Runners.
Under "Runners", you can view a list of registered runners, including the runner's name, labels, and status.

The status can be one of the following:
- Idle: The runner is connected to GitHub Enterprise Cloud and is ready to execute jobs.
- Active: The runner is currently executing a job.
- Offline: The runner is not connected to GitHub Enterprise Cloud. This could be because the machine is offline, the self-hosted runner application is not running on the machine, or the self-hosted runner application cannot communicate with GitHub Enterprise Cloud.

Troubleshooting network connectivity

Checking self-hosted runner network connectivity

You can use the self-hosted runner application's config script with the --check parameter to check that a self-hosted runner can access all required network services on GitHub.

In addition to --check, you must provide two arguments to the script:

--url with the URL to your GitHub repository, organization, or enterprise. For example, --url https://github.com/octo-org/octo-repo.
--pat with the value of a personal access token (classic), which must have the workflow scope, or a fine-grained personal access token with workflows read and write access. For example, --pat ghp_abcd1234. For more information, see "Managing your personal access tokens."

For example:

./config.sh --check --url URL --pat ghp_abcd1234

./config.sh --check --url URL --pat ghp_abcd1234

config.cmd --check --url https://github.com/YOUR-ORG/YOUR-REPO --pat GHP_ABCD1234

The script tests each service, and outputs either a PASS or FAIL for each one. If you have any failing checks, you can see more details on the problem in the log file for the check. The log files are located in the _diag directory where you installed the runner application, and the path of the log file for each check is shown in the console output of the script.

If you have any failing checks, you should also verify that your self-hosted runner machine meets all the communication requirements. For more information, see "About self-hosted runners."

Disabling TLS certificate verification

By default, the self-hosted runner application verifies the TLS certificate for GitHub Enterprise Cloud. If you encounter network problems, you may wish to disable TLS certificate verification for testing purposes.

To disable TLS certification verification in the self-hosted runner application, set the GITHUB_ACTIONS_RUNNER_TLS_NO_VERIFY environment variable to 1 before configuring and running the self-hosted runner application.

export GITHUB_ACTIONS_RUNNER_TLS_NO_VERIFY=1
./config.sh --url https://github.com/YOUR-ORG/YOUR-REPO --token
./run.sh

export GITHUB_ACTIONS_RUNNER_TLS_NO_VERIFY=1
./config.sh --url https://github.com/YOUR-ORG/YOUR-REPO --token
./run.sh

[Environment]::SetEnvironmentVariable('GITHUB_ACTIONS_RUNNER_TLS_NO_VERIFY', '1')
./config.cmd --url https://github.com/YOUR-ORG/YOUR-REPO --token
./run.cmd

Warning

Disabling TLS verification is not recommended since TLS provides privacy and data integrity between the self-hosted runner application and GitHub Enterprise Cloud. We recommend that you install the GitHub Enterprise Cloud certificate in the operating system certificate store for your self-hosted runner. For guidance on how to install the GitHub Enterprise Cloud certificate, check with your operating system vendor.

Reviewing the self-hosted runner application log files

You can monitor the status of the self-hosted runner application and its activities. Log files are kept in the _diag directory where you installed the runner application, and a new log is generated each time the application is started. The filename begins with Runner_, and is followed by a UTC timestamp of when the application was started.

Warning

Runner application log files for ephemeral runners must be forwarded and preserved externally for troubleshooting and diagnostic purposes. For more information about ephemeral runners and autoscaling self-hosted runners, see "Autoscaling with self-hosted runners."

For detailed logs on workflow job executions, see the next section describing the Worker_ files.

Reviewing a job's log file

The self-hosted runner application creates a detailed log file for each job that it processes. These files are stored in the _diag directory where you installed the runner application, and the filename begins with Worker_.

Using journalctl to check the self-hosted runner application service

For Linux-based self-hosted runners running the application using a service, you can use journalctl to monitor their real-time activity. The default systemd-based service uses the following naming convention: actions.runner.<org>-<repo>.<runnerName>.service. This name is truncated if it exceeds 80 characters, so the preferred way of finding the service's name is by checking the .service file. For example:

$ cat ~/actions-runner/.service
actions.runner.octo-org-octo-repo.runner01.service

If this fails due to the service being installed elsewhere, you can find the service name in the list of running services. For example, on most Linux systems you can use the systemctl command:

$ systemctl --type=service | grep actions.runner
actions.runner.octo-org-octo-repo.hostname.service loaded active running GitHub Actions Runner (octo-org-octo-repo.hostname)

You can use journalctl to monitor the real-time activity of the self-hosted runner:

sudo journalctl -u actions.runner.octo-org-octo-repo.runner01.service -f

In this example output, you can see runner01 start, receive a job named testAction, and then display the resulting status:

Feb 11 14:57:07 runner01 runsvc.sh[962]: Starting Runner listener with startup type: service
Feb 11 14:57:07 runner01 runsvc.sh[962]: Started listener process
Feb 11 14:57:07 runner01 runsvc.sh[962]: Started running service
Feb 11 14:57:16 runner01 runsvc.sh[962]: √ Connected to GitHub
Feb 11 14:57:17 runner01 runsvc.sh[962]: 2020-02-11 14:57:17Z: Listening for Jobs
Feb 11 16:06:54 runner01 runsvc.sh[962]: 2020-02-11 16:06:54Z: Running job: testAction
Feb 11 16:07:10 runner01 runsvc.sh[962]: 2020-02-11 16:07:10Z: Job testAction completed with result: Succeeded

To view the systemd configuration, you can locate the service file here: /etc/systemd/system/actions.runner.<org>-<repo>.<runnerName>.service. If you want to customize the self-hosted runner application service, do not directly modify this file. Follow the instructions described in "Configuring the self-hosted runner application as a service."

Using `launchd` to check the self-hosted runner application service

For macOS-based self-hosted runners running the application as a service, you can use launchctl to monitor their real-time activity. The default launchd-based service uses the following naming convention: actions.runner.<org>-<repo>.<runnerName>. This name is truncated if it exceeds 80 characters, so the preferred way of finding the service's name is by checking the .service file in the runner directory:

% cat ~/actions-runner/.service
/Users/exampleUsername/Library/LaunchAgents/actions.runner.octo-org-octo-repo.runner01.plist

The svc.sh script uses launchctl to check whether the application is running. For example:

$ ./svc.sh status
status actions.runner.example.runner01:
/Users/exampleUsername/Library/LaunchAgents/actions.runner.example.runner01.plist
Started:
379 0 actions.runner.example.runner01

The resulting output includes the process ID and the name of the application’s launchd service.

To view the launchd configuration, you can locate the service file here: /Users/exampleUsername/Library/LaunchAgents/actions.runner.<repoName>.<runnerName>.service. If you want to customize the self-hosted runner application service, do not directly modify this file. Follow the instructions described in "Configuring the self-hosted runner application as a service."

Using PowerShell to check the self-hosted runner application service

For Windows-based self-hosted runners running the application as a service, you can use PowerShell to monitor their real-time activity. The service uses the naming convention GitHub Actions Runner (<org>-<repo>.<runnerName>). You can also find the service's name by checking the .service file in the runner directory:

PS C:\actions-runner> Get-Content .service
actions.runner.octo-org-octo-repo.runner01.service

You can view the status of the runner in the Windows Services application (services.msc). You can also use PowerShell to check whether the service is running:

PS C:\actions-runner> Get-Service "actions.runner.octo-org-octo-repo.runner01.service" | Select-Object Name, Status
Name                                                  Status
----                                                  ------
actions.runner.octo-org-octo-repo.runner01.service    Running

You can use PowerShell to check the recent activity of the self-hosted runner. In this example output, you can see the application start, receive a job named testAction, and then display the resulting status:

PS C:\actions-runner> Get-EventLog -LogName Application -Source ActionsRunnerService

   Index Time          EntryType   Source                 InstanceID Message
   ----- ----          ---------   ------                 ---------- -------
     136 Mar 17 13:45  Information ActionsRunnerService          100 2020-03-17 13:45:48Z: Job Greeting completed with result: Succeeded
     135 Mar 17 13:45  Information ActionsRunnerService          100 2020-03-17 13:45:34Z: Running job: testAction
     134 Mar 17 13:41  Information ActionsRunnerService          100 2020-03-17 13:41:54Z: Listening for Jobs
     133 Mar 17 13:41  Information ActionsRunnerService          100 û Connected to GitHub
     132 Mar 17 13:41  Information ActionsRunnerService            0 Service started successfully.
     131 Mar 17 13:41  Information ActionsRunnerService          100 Starting Actions Runner listener
     130 Mar 17 13:41  Information ActionsRunnerService          100 Starting Actions Runner Service
     129 Mar 17 13:41  Information ActionsRunnerService          100 create event log trace source for actions-runner service

Monitoring the automatic update process

We recommend that you regularly check the automatic update process, as the self-hosted runner will not be able to process jobs if it falls below a certain version threshold. The self-hosted runner application automatically updates itself, but note that this process does not include any updates to the operating system or other software; you will need to separately manage these updates.

You can view the update activities in the Runner_ log files. For example:

[Feb 12 12:37:07 INFO SelfUpdater] An update is available.

In addition, you can find more information in the SelfUpdate log files located in the _diag directory where you installed the runner application.

Troubleshooting containers in self-hosted runners

Checking that Docker is installed

If your jobs require containers, then the self-hosted runner must be Linux-based and needs to have Docker installed. Check that your self-hosted runner has Docker installed and that the service is running.

You can use systemctl to check the service status:

$ sudo systemctl is-active docker.service
active

If Docker is not installed, then dependent actions will fail with the following errors:

[2020-02-13 16:56:10Z INFO DockerCommandManager] Which: 'docker'
[2020-02-13 16:56:10Z INFO DockerCommandManager] Not found.
[2020-02-13 16:56:10Z ERR  StepsRunner] Caught exception from step: System.IO.FileNotFoundException: File not found: 'docker'

Checking the Docker permissions

If your job fails with the following error:

dial unix /var/run/docker.sock: connect: permission denied

Check that the self-hosted runner's service account has permission to use the Docker service. You can identify this account by checking the configuration of the self-hosted runner in systemd. For example:

$ sudo systemctl show -p User actions.runner.octo-org-octo-repo.runner01.service
User=runner-user

Checking which Docker engine is installed on the runner

If your build fails with the following error:

Error: Input required and not supplied: java-version

Check which Docker engine is installed on your self-hosted runner. To pass the inputs of an action into the Docker container, the runner uses environment variables that might contain dashes as part of their names. The action may not be able to get the inputs if the Docker engine is not a binary executable, but is instead a shell wrapper or a link (for example, a Docker engine installed on Linux using snap). To address this error, configure your self-hosted runner to use a different Docker engine.

To check if your Docker engine was installed using snap, use the which command. In the following example, the Docker engine was installed using snap:

$ which docker
/snap/bin/docker