GH-100 · Domain 4: Manage GitHub Actions · 24% of the exam
Task 4.2: Manage runners
Runner groups, GitHub-hosted, larger and self-hosted runners, IP allow lists and Azure private networking, and monitoring and troubleshooting runners.
Study it
Runner groups, larger and self-hosted runners, IP allow lists and Azure private networking
Lesson coming
Sample questions
Try each one before opening the answer. Every option is explained, with the GitHub documentation page that proves it.
Question 1 · choose 1
An enterprise owner creates the enterprise runner group gpu-large for larger runners with GPUs and sets its organization access to the ml-org organization only. A week later, jobs in the private repository ml-org/train still can't use the group's runners, although every repository of ml-org can use ml-org's own runner groups. What has to happen for ml-org/train?
AThe enterprise owner adds ml-org/train to the repository access of gpu-large in the enterprise settings
BNothing more; the job needs a runner label, because sharing gpu-large with ml-org gave all its repositories access
CAn owner of ml-org configures gpu-large in the organization's settings to give ml-org/train access
DThe enterprise owner changes the organization access of gpu-large to all organizations, which opens it to every repository
Show the answer and why
AThe enterprise owner adds ml-org/train to the repository access of gpu-large in the enterprise settings
Incorrect
At the enterprise level a runner group's access names organizations; organization owners choose which of their repositories get access.
BNothing more; the job needs a runner label, because sharing gpu-large with ml-org gave all its repositories access
Incorrect
Repositories don't get access to an enterprise runner group by default, unlike organization runner groups.
CAn owner of ml-org configures gpu-large in the organization's settings to give ml-org/train access
Correct
After the enterprise shares the group, organization owners must choose which repositories can use it.
DThe enterprise owner changes the organization access of gpu-large to all organizations, which opens it to every repository
Incorrect
Organization access alone still leaves the repositories without access until an organization owner grants it.
Access to runners follows the levels where the runner group is defined. An enterprise runner group is shared with organizations at the enterprise level, and by default the repositories in an organization don't have access to enterprise runner groups: organization owners must configure each shared group and choose the repositories. Organization runner groups work the other way, granting all repositories in the organization access by default.
A platform team will run autoscaling self-hosted runners on its OpenShift cluster with Actions Runner Controller (ARC), and it wants GitHub Support to be able to help with ARC problems. Which deployment fits?
AThe latest version of GitHub's Autoscaling Runner Sets, installed with Helm
BThe legacy community-maintained ARC, installed with Helm on the cluster
CGitHub's Autoscaling Runner Sets, installed with Kustomize and a customized template spec
DAutoscaling Runner Sets on a separate Kubernetes cluster, because OpenShift isn't supported
Show the answer and why
AThe latest version of GitHub's Autoscaling Runner Sets, installed with Helm
Correct
GitHub supports only the latest Autoscaling Runner Sets version, and tooling other than Helm is out of scope.
BThe legacy community-maintained ARC, installed with Helm on the cluster
Incorrect
Support for the legacy ARC comes from the community only.
CGitHub's Autoscaling Runner Sets, installed with Kustomize and a customized template spec
Incorrect
Installation tooling other than Helm and template spec customization fall outside GitHub Support's scope.
DAutoscaling Runner Sets on a separate Kubernetes cluster, because OpenShift isn't supported
Incorrect
GitHub's Autoscaling Runner Sets work with both Kubernetes and OpenShift.
There are two ARC releases: the legacy community-maintained ARC and GitHub's Autoscaling Runner Sets, which work with Kubernetes and OpenShift. GitHub only supports the latest Autoscaling Runner Sets version. Requests about deployments with installation tooling other than Helm, or with customized service accounts or template specs, may fall outside GitHub Support's scope.
An enterprise turns on its IP allow list with the office ranges. Its self-hosted runners run in a cloud network and reach GitHub through a NAT gateway whose public address is 203.0.113.10. Since the change, their jobs fail at the actions/checkout step. What should an enterprise owner do?
ANothing on GitHub; the network team should open outbound HTTPS from the cloud network
BAdd 203.0.113.10 to the enterprise's IP allow list
CHave the jobs use a personal access token stored as a secret instead of the GITHUB_TOKEN
DSwitch the checkout step to SSH with a deploy key added to each repository
Show the answer and why
ANothing on GitHub; the network team should open outbound HTTPS from the cloud network
Incorrect
GitHub refuses the requests because the allow list also applies to the job's GITHUB_TOKEN.
BAdd 203.0.113.10 to the enterprise's IP allow list
Correct
The list applies to every authentication method used from an address, including the GITHUB_TOKEN.
CHave the jobs use a personal access token stored as a secret instead of the GITHUB_TOKEN
Incorrect
The IP allow list restricts personal access tokens as well.
DSwitch the checkout step to SSH with a deploy key added to each repository
Incorrect
The IP allow list also restricts SSH keys, deploy keys included.
Once an IP allow list is enabled, it decides whether users or apps can reach protected resources through the web UI, APIs or Git, for interactive sign-in and for personal access tokens, SSH keys including deploy keys, and GitHub App tokens including the GitHub Actions GITHUB_TOKEN. Requests from the runners' public address must therefore be allowed.
After a proxy change, every runner in an enterprise-level self-hosted runner group shows as Offline. Before changing the proxy again, the enterprise owner wants to find out, from one runner machine, which required GitHub services it can't reach. What should she run on that machine?
ASet GITHUB_ACTIONS_RUNNER_TLS_NO_VERIFY to 1 and restart the runner application
B./config.sh --check with the enterprise URL and a fine-grained token with read-only Actions access
CA re-run of a failed job with debug logging, to read the network errors
D./config.sh --check with --url set to the enterprise URL and --pat set to a classic token with the workflow scope
Show the answer and why
ASet GITHUB_ACTIONS_RUNNER_TLS_NO_VERIFY to 1 and restart the runner application
Incorrect
That disables TLS verification for testing; it doesn't report which services are reachable.
B./config.sh --check with the enterprise URL and a fine-grained token with read-only Actions access
Incorrect
The check needs a classic token with the workflow scope, or a fine-grained token with read and write access to workflows.
CA re-run of a failed job with debug logging, to read the network errors
Incorrect
An offline runner isn't connected to GitHub, so it can't pick up the job.
D./config.sh --check with --url set to the enterprise URL and --pat set to a classic token with the workflow scope
Correct
The check tests each required GitHub service from the machine and reports PASS or FAIL for each.
The runner application's config script with --check tests whether a self-hosted runner can reach all required network services on GitHub. It needs --url with the repository, organization or enterprise URL, and --pat with a personal access token (classic) that has the workflow scope, or a fine-grained token with read and write access to workflows.
Jobs on GitHub-hosted runners must call a service in a company's on-premises network. The company has no Azure subscription and doesn't want to run self-hosted runners. Which two approaches does GitHub describe? (Choose TWO.)
AAdd the on-premises service's address to the enterprise's IP allow list
BHave the jobs authenticate to the service by sending the GITHUB_TOKEN in a header
CAn API gateway in front of the service that accepts the job's OIDC token
DAn environment with required reviewers on the jobs that call the service
EA WireGuard overlay network between the runner and a host in the private network
Show the answer and why
AAdd the on-premises service's address to the enterprise's IP allow list
Incorrect
The IP allow list controls access to the enterprise's resources on GitHub, not runners' access to other networks.
BHave the jobs authenticate to the service by sending the GITHUB_TOKEN in a header
Incorrect
The gateway approach authenticates workflows outside GitHub with OIDC tokens, not the GITHUB_TOKEN.
CAn API gateway in front of the service that accepts the job's OIDC token
Correct
Workflows can authenticate outside GitHub Actions with OIDC tokens through an API gateway.
DAn environment with required reviewers on the jobs that call the service
Incorrect
Required reviewers decide when a job may run; they don't connect it to a private network.
EA WireGuard overlay network between the runner and a host in the private network
Correct
Running WireGuard in both places creates an overlay network to the private service.
GitHub-hosted runners have access to the public internet by default. To reach resources on a private network, workflows can authenticate to an API gateway with OpenID Connect tokens, or a WireGuard overlay network can link the runner with a service in the private network, without separate gateway infrastructure.
An organization owner tries to create an organization-level runner group named linux-builders, but GitHub doesn't allow it. The enterprise already has a runner group called linux-builders that is shared with the organization. Why?
AOrganizations can't create their own runner groups at all once an enterprise group is shared with them
BRunner group names must be unique at the organization level, enterprise groups included
COnly enterprise owners can create runner groups in enterprise organizations
DThe name must start with the organization's name
Show the answer and why
AOrganizations can't create their own runner groups at all once an enterprise group is shared with them
Incorrect
Organizations can still create runner groups, just not with a name already in use.
BRunner group names must be unique at the organization level, enterprise groups included
Correct
An organization can't create a group with the same name as an enterprise group.
COnly enterprise owners can create runner groups in enterprise organizations
Incorrect
Organization owners can create organization-level runner groups.
DThe name must start with the organization's name
Incorrect
There's no prefix rule; names must be unique.
GitHub Actions requires runner group names to be unique at the organization level, so an organization can't create a runner group with the same name as one in the enterprise. Runner groups that already share a name show a warning banner.
An enterprise with managed users uses its identity provider's allow list, enforced through the IdP's Conditional Access Policy, instead of GitHub's IP allow list. For which requests is the IdP's policy not applied?
AWeb requests from members signing in
BAPI requests with a user's personal access token
CRequests from a GitHub App using an installation access token
DAPI requests made by an OAuth app that is acting on behalf of a user
Show the answer and why
AWeb requests from members signing in
Incorrect
The OIDC CAP applies to web requests.
BAPI requests with a user's personal access token
Incorrect
The CAP applies to API requests made with user tokens.
CRequests from a GitHub App using an installation access token
Correct
The OIDC CAP doesn't apply when a GitHub App uses an installation access token.
DAPI requests made by an OAuth app that is acting on behalf of a user
Incorrect
The CAP applies to user tokens such as OAuth app tokens.
When an enterprise uses its IdP's allow list instead of GitHub's, the IdP's OIDC Conditional Access Policy applies to web requests and to API requests with user tokens, such as OAuth app tokens or GitHub App user access tokens, but not to GitHub App installation access tokens.
An enterprise owner wants a list of all jobs running right now on GitHub-hosted runners across the enterprise. Where does she find it?
AIn the enterprise's Actions policies, Runners tab, under GitHub-hosted runners and then Active jobs
BIn the enterprise audit log, filtered to workflow events
CIn each repository's Insights tab
DIn the billing usage report for the current day
Show the answer and why
AIn the enterprise's Actions policies, Runners tab, under GitHub-hosted runners and then Active jobs
Correct
The GitHub-hosted runners entry lists every job currently running on them.
BIn the enterprise audit log, filtered to workflow events
Incorrect
The audit log records events, not a live list of running jobs.
CIn each repository's Insights tab
Incorrect
Repository insights don't list jobs running across the enterprise.
DIn the billing usage report for the current day
Incorrect
Usage reports summarize spend; they don't show running jobs.
Organization owners and enterprise owners can see all jobs currently running on GitHub-hosted runners. For an enterprise, open Policies, then Actions, then the Runners tab, select the GitHub-hosted runners entry, and review the Active jobs section.