Microsoft’s code internet hosting shack Github has revealed a prolonged mea culpa about its availability and reliability woes – one that features the phrases “we’re sorry.”
GitHub seems to be fighting measly three nines availability
READ MORE
Builders utilizing the service have confronted repeated downtime, and the apology comes days after Hashicorp co-founder Mitchell Hashimoto declared the platform is “now not a spot for severe work”, asserting plans to maneuver his terminal emulator mission, Ghostty, elsewhere.
GitHub highlighted a number of latest incidents in its submit. On April 23th, a Merge Queue bug brought on merge teams containing a couple of pull request to provide incorrect commits.
“In affected circumstances,” notes GitHub, “modifications from beforehand merged pull requests and prior commits have been inadvertently reverted by subsequent merges.”
Then on April twenty seventh, elements of the person interface depending on search stopped displaying outcomes after GitHub’s Elasticsearch cluster grew to become “overloaded (doubtless as a consequence of a botnet assault).” GitHub remains to be engaged on a root trigger evaluation.
The issues are usually not new. In February, The Register famous that GitHub was fighting availability. Uptime for the service, in keeping with a reconstructed status page, dropped under 90 p.c in 2025, and the pattern has been downward. April’s uptime is now under 85 p.c.
AI is an element, in keeping with the supply shack: “The principle driver is a speedy change in how software program is being constructed. For the reason that second half of December 2025, agentic growth workflows have accelerated sharply.”
In between the handwringing, GitHub admits it had deliberate to up capability by 10X and mentioned it started work to do that in October 2025. Nonetheless, by February this 12 months, it grew to become clear that 30 instances as a lot capability is required. GitHub has since scrambled to cope with the problems.
“Our priorities are clear: availability first, then capability, then new options. We’re lowering pointless work, bettering caching, isolating vital providers, eradicating single factors of failure, and transferring performance-sensitive paths into programs designed for these workloads.”
What concerning the migration to Azure? That’s not the offender. GitHub says the migration has truly helped, permitting it to “arise much more compute” shortly.
Yesterday, Hashimoto complained in a post on his website:
“For the previous month I’ve stored a journal the place I put an ‘X’ subsequent to each date the place a GitHub outage has negatively impacted my capacity to work. Nearly day by day has an ‘X’. On the day I’m penning this submit, I have been unable to do any PR evaluate for ~2 hours as a result of there’s a GitHub Actions outage.”
GitHub’s apology was penned by Chief Know-how Officer, Vlad Fedorov, who closed with: “We hear the ache you are experiencing. We learn each electronic mail, social submit, help ticket, and we take all of it to coronary heart.”
“We’re sorry.”
Restoring developer confidence to pre-crisis ranges – when it was the go-to place for stashing code and collaborating on initiatives – will take greater than phrases.
®