Incident with Pull Requests
Affected
Operational from 11:59 AM to 3:33 PM, Degraded performance from 3:33 PM to 4:07 PM, Operational from 4:07 PM to 4:26 PM
Operational from 11:59 AM to 3:33 PM, Degraded performance from 3:33 PM to 4:07 PM, Operational from 4:07 PM to 4:26 PM
- Resolved · UTC
On November 16, 2026, between 09:30 and 15:55 UTC, some pull request merges on GitHub timed out. On average, 2.9% of merge attempts timed out, with a peak of 7.1% at 15:17 UTC. Some GitHub Actions workflow runs triggered by pull requests were also delayed. Git operations such as push and fetch stayed available, with higher latency. This was a repeat of a shorter event on November 13, between 13:15 and 17:50 UTC, when 1.7% of merges timed out.
Both events had the same cause: some of the storage servers that hold Git repository data reached their disk read bandwidth limits. A merge must be written to every copy of a repository, so one overloaded server slowed merges for every repository it hosts. Read load on these servers had been growing as part of a planned move of repositories to new infrastructure. A disk configuration setting on those servers also made them read far more data than necessary, which cut their usable bandwidth roughly in half. We mitigated the incident by correcting that setting and moving some load away from the busiest servers.
The corrected setting is now applied permanently across our storage fleet. We are adding storage capacity in the affected region and alerting on per-server disk bandwidth so we catch saturation before it affects customers. Sustained merge timeouts will now page our on-call engineers directly. We are also adding capacity checks before moving repositories or traffic onto new infrastructure. - Update · UTC
We’ve mitigated the issue affecting pull request merges and Actions workflow start times by reducing load on affected systems. Pull Requests and Actions have recovered, and we’re continuing to monitor their performance.
- Monitoring · UTC
The degradation affecting Actions and Pull Requests has been mitigated. We are monitoring to ensure stability.
- Update · UTC
We’re seeing early signs of improvement in pull request merge times & workflow start times, but recovery is not yet confirmed. Users may still have difficulty merging pull requests. We’re reducing load on affected systems and monitoring the impact. We’ll share another update as we learn more.
- Update · UTC
We’re seeing early signs of improvement in pull request merge times, but recovery is not yet confirmed. The last confirmed timeout rate was 2.5%, compared with a baseline near 0.2%, so some users may still have difficulty merging pull requests. We’ll share another update as we learn more.
- Investigating · UTC
Actions is experiencing degraded performance. We are continuing to investigate.

