Fair enough; since fork-ci is opt-in, it's better to watch the impact of potential GitHub restrictions. This was recently released by ASF Infra. Let's look at the results of how other PMCs are utilising this. We need to run CIs before merging anyway.
On Sun, Nov 22, 2026 at 10:26 PM Ádám Sághy <[email protected]> wrote: > @Krishna: Could you share details about the reasons behind the GitHub > account restriction, if there is any update? > > > If extensive GitHub Actions use led to this, we should proceed cautiously > before requiring contributors to use their own forks for those runs. > > > I am unsure whether this step is necessary. We could schedule them during > off-peak periods and withhold CI approval when the queue is full. > > > Regards, > > Adam S > > On Nov 22, 2026, at 4:47 PM, Aman Mittal <[email protected]> > wrote: > > Hi Victor, > > I don't think CI runners on CI is slow, as both the ASF repo and forked > repos use standard runners. The ASF enterprise runners have shared large > pools and caching support compared to the 20 concurrent runners on > individual forks. They don't need manual approval to trigger runs. It will > be faster if we manually approve CIs after each push. I think i should be > explicit in my wording that it will not replace the repo actions but will > reduce the need for manual approvals in this. > > On Sun, Nov 22, 2026 at 10:05 PM VICTOR MANUEL ROMERO RODRIGUEZ < > [email protected]> wrote: > >> -1 they are slower >> >> El dom., 22 de noviembre de 2026 10:24 a. m., Aman Mittal < >> [email protected]> escribió: >> >>> Hi All, >>> >>> I'd like to propose adopting ASF Infra's *fork-ci* action ( >>> apache/infrastructure-actions/fork-ci) to save shared runner minutes >>> and cut down on clicking *"Approve and run"* for external PRs. >>> How it works >>> >>> - >>> >>> *Offloads WIP runs to forks:* Contributors run our CI workflow on >>> their own public forks (using their own GitHub Actions quota) when they >>> push commits. >>> - >>> >>> *Mirrors results automatically:* A lightweight bridge + 15-minute >>> cron sweep post the fork's run status (CI (fork)) directly onto the >>> Fineract PR including links and failed job summaries even if the upstream >>> workflow hasn't been approved yet. >>> - >>> >>> *Safe for merging:* Since fork runs are untrusted, CI (fork) is *strictly >>> informational* (the bridge never checks out PR code or touches >>> secrets). Full, trusted upstream CI still gates every merge via >>> merge_group / committer approval. >>> >>> Rollout Plan >>> >>> 1. >>> >>> *Phase 1 (Informational only):* Add fork-ci-signal.yml and >>> fork-ci-status.yml to develop (pinned to a reviewed SHA) without >>> skipping any existing upstream checks yet. >>> 2. >>> >>> *Phase 2 (Docs):* Add a quick note in our contributor guide on >>> enabling GitHub Actions on forks (guide-url). >>> 3. >>> >>> *Phase 3 (Optimize):* Once verified alongside full merge_group >>> validation, skip redundant heavy upstream jobs on initial external >>> PR pushes. >>> >>> Docs: >>> https://github.com/apache/infrastructure-actions/blob/main/fork-ci/README.md >>> >>> Let me know what you think. if there are no objections, I'll file a >>> JIRA and open a PR for Phase 1. >>> >>> Thanks, >>> >>> Aman >>> >>> >
