Public REST & GraphQL gateway
Container scheduling & autoscale pool
S3-compatible buckets, all regions
Managed primaries and read replicas
console.halyard.dev web app
CI images, cache layer, artifact upload
Package downloads and version manifest
Reference site and semantic index
Primary North America region
EU residency zone
APAC traffic and replication
Cache hit rates are back above 92%. We are leaving the incident open while the warm pool refills.
A bad eviction policy shipped with runner image 4.19 caused the shared layer cache to thrash. Rolling back to 4.18.
Builds are queuing 3–6× longer than baseline in us-east-1. Investigating.
Reindex finished with no data loss. Search latency improved from 240ms to 96ms at p95.
Search on docs.halyard.dev falls back to keyword matching for the duration of the window.
All upload paths healthy for 20 consecutive minutes. A full write-up is published in our postmortem archive.
A metadata shard failed to rejoin quorum after a routine host replacement. Draining traffic to the remaining two shards.
Multipart uploads over 8 MB are failing in Singapore. Single-part uploads and all reads are unaffected.
Scale-up requests are succeeding normally. We have added a rate limit to the placement controller.
Roughly 2% of scale-up requests are returning a placement error. Existing workloads are not affected.
One email per incident, plus a weekly digest. Webhook, Slack and RSS delivery are available from the same preferences page.