No visible pull_request runs; merge ref returns 404 in a private repository #212773
Replies: 1 comment 5 replies
|
I can't see your repository, so this is general guidance from the documented behavior, not a diagnosis. Please verify each point against the current docs before acting. The merge ref. Whether the event was received. I'm not aware of a UI or API view that shows event delivery for workflows that never started. The closest checks are the Actions run list filtered by Billing versus budget. A configured spending budget is not the same as an active execution block. Included-minute exhaustion and a $0 spending limit can both stop jobs, but a budget set above zero should allow paid usage. Check the Billing and licensing page for the exact limit and any "payment" warnings, and remember that a run can stay queued for a while after a limit changes. Minimal evidence before changing anything: the workflow file at the head commit, the |
Uh oh!
There was an error while loading. Please reload this page.
🏷️ Discussion Type
Question
💬 Feature/Topic Area
Workflow Configuration
Discussion Details
Environment
I am investigating an open draft PR in a private repository. The repository owner reports a personal GitHub Free account. The PR's source and target branches are in the same repository. The workflow is configured to use GitHub-hosted runners.
I am looking for read-only ways to distinguish possible causes, not asking anyone to diagnose a private repository from this description. Repository names, IDs, branch names, commit hashes, and exact timestamps are intentionally omitted.
Observed behavior
In the latest authenticated read-only snapshot:
Actions run queries filtered separately by the current PR head SHA and source branch returned zero runs, without an event-type filter.
The head commit had zero check runs and zero individual commit statuses. The combined status response said pending, but its status list was empty; I am not treating that as evidence of a queued Actions run.
A request for refs/pull/<PR_NUMBER>/merge returned HTTP 404.
The raw PR response reported mergeable: null, mergeable_state: "unknown", and no merge commit SHA. The PR remained open and draft.
These are observations, not proof that the event was never received, that a merge conflict exists, or that the 404 caused the missing runs.
Checks already performed
The workflow is also present at the inspected base commit. At the inspected head it subscribes to pull_request without activity-type, path, or branch filters. The documented default activity types are opened, synchronize, and reopened. Its separate push trigger is restricted to the default branch, so a feature-branch push alone is not expected to trigger that path. I found no job-level condition excluding draft PRs.
An earlier retained API snapshot showed the workflow active and repository Actions enabled. These settings were not freshly revalidated in the latest snapshot because the available read-only client does not support those endpoints. The earlier snapshot and current observations are kept separate.
The account owner reports that the included Actions allowance was exhausted and that an Actions spending budget was subsequently configured. Current payment eligibility and whether billing blocked these particular events have not been verified. An Actions budget is not a GitHub Pro subscription.
I have read the pull_request event documentation and the PR REST documentation. I understand that a merge conflict can prevent this event's workflows from running, and that mergeable: null can mean computation is pending. Neither establishes the cause here.
Questions
Besides a merge conflict, what documented conditions can leave an open PR's merge ref unavailable? Which read-only checks distinguish pending mergeability computation, access limitations, and other conditions without inferring a conflict from 404 alone?
Is there a supported read-only UI or API view that distinguishes event receipt, filtering, approval requirements, billing restrictions, and queueing when the run list is empty? Which stages are not visible to repository owners? Which non-secret details about how the PR was created or updated should be collected?
For this trigger configuration, how should draft status, GitHub Free allowances, payment eligibility, and Actions budget limits be evaluated separately? Which read-only billing indicators establish an actual execution block rather than merely a configured budget?
What minimal sanitized evidence should be collected before changing the PR, workflow, or repository settings?
Actions not performed
During this diagnostic pass I have not reopened the PR, marked it ready, dispatched a workflow, changed Actions settings or billing, or merged anything to provoke a run. I would prefer to preserve the current state while collecting evidence.
Additional sanitized evidence available upon request
I can prepare a generalized trigger fragment and a redacted observation table for review. I will not publish raw API responses, credentials, private repository identifiers, or private logs. Any additional disclosure will be reviewed first.
I found related discussions #206064 and #210548, both marked unanswered when checked. The former also involves self-hosted runners and stuck runs; the latter involves a newly added workflow and little allowance usage. I have not established that either shares this cause. If this belongs in an existing discussion, please point me to it.
All reactions