September was a month of infrastructure trouble, looking back, and fixes. Plan upgrades now change your app’s limits right away, deleting apps no longer messes with your quota, and the CLI got a bit sturdier.
Early in September we ran a series of upgrades on our clusters. As things go, staging worked beautifully — but one of the upgrades on production failed and caused almost two hours of downtime. We’re sorry about that.
Afterwards, we took a close look at all our setups to find out why production failed when staging didn’t. We have a strong lead, though there’s no way to be fully sure without trying again. In short: some of the versions running underneath production had drifted from staging, and the upgrade failed because of that.
We’ve since planned the next steps: update the drifted components and rebuild parts of the infrastructure. This includes changes to the control plane, and we’ll roll them out step by step — with notice on our status page where needed.
On top of that, we spent a good part of the month on problems with a vendor’s backup and storage product. Nobody on the team would call that fun: it’s closed-source software, so we depend on their support to move things forward.
We’re planning work for early 2027 to replace it. With hardware prices going up almost every day, that’s a real challenge for a small business like ours — but we think it’s the right call.
When you moved an app to a bigger plan, the new limits didn’t always reach the running app — it could stay on the memory limit of its old plan, even after a redeploy. That’s fixed:
runway app plan set --plan=<plan>
Your app now gets the resources of its new plan. See the plans reference for all options.
Deleting an app that was already being deleted counted against your quota a second time. In the worst case, this blocked later plan upgrades. Thanks for the report!
We fixed the cause and checked every account — only one needed a correction, and that’s done.
We shipped a few updates to the Runway CLI. For example, runway login no longer crashes when the network has a hiccup — you get a proper error instead. And runway app config commands now tell you more clearly what the API did.
Tuesday (20th Oct) we did another release (v1.48.1) rebuilt with an updated Go version to deal with the eMudhra certificate leak.
The rest of our services have had this rebuild as well.
Update with brew upgrade or see the install docs.
Node.js and PHP both shipped security releases this week, fixing a long list of CVEs. We updated all included buildpacks on 2026-10-21 to pick them up, and the docs now list the new versions.
Your running apps keep the runtime they were built with. To get the patched Node.js or PHP versions, deploy your app again:
git commit --allow-empty -m "chore: rebuild for security updates"
runway app deploy
If you pin an exact version (for example in package.json under engines, or in composer.json), update the pin first, otherwise the build keeps the old version. A version range like 22.x or ^8.4 picks up the patch on its own.
Further reading:
That’s it for September. Months like this make it easy to lose drive — but that’s not us. We’re keeping this one short, because October is going a lot better, and we hope to have a great update for you at the end of the month!
As always, let us know what you think — and if you’re new here: Runway runs your apps and managed databases, hosted entirely in the EU.