Using Kubernetes Event Driver Autoscaling to scale worker pods to 0 based on reading queues or checking APIs. Implemented for Pixelfed, Bookwyrm, and gitea action runners, with the runners scaling up to 4 pods, with examples.
I’m pretty happy with the progress that I’ve made in the past few days. I’m hosting Keyboard Vagabond, a collection of Fediverse apps for the traveler/digital nomad niche, on a set of VPS’s in a self-hosted kubernetes cluster.
Changing Piefed Worker Scaling to be Based on Queue Size in Kubernetes with KEDA
I recently caused myself a bit of a minor issue by installing some updates on the Keyboard Vagabond cluster. It wasn’t a big deal, just s…
It started with a perfectly good and running kubernetes cluster hosting fediverse applications at keyboardvagabond with all the infrastructure and observability that comes with it. Then I made a simple networking fix.
Moar adventures in #selfhosting for #keyboardvagabond! I think that I finally got the longhorn too-many-s3-calls networking issue resolved blog.keyboardvagabond.com/mich
Usually I post updates like these on my gotosocial account, but my computer/server is at my parents house and their modem has been having a moment for the past day and a half and they're not the best sys-admins. I have more posts and updates that would normally be found on mastodon, but again - parents modem haha.