All deploys to our LHR-region apps have been failing for ~2 hours. Builds complete successfully but the push to the internal registry fails every time:
Attempts: 10+ deploys, tried --remote-only, local Docker build, --depot=false, destroyed and recreated builder — all fail at the same registry push step
Status page shows “All Systems Operational” but registry node fdaa::3 is clearly down
Last successful deploys were yesterday (25 March). Existing apps are running fine — just can’t push new images.
Tried the suggestions — --recreate-builder and --depot=false. The deploy cycled through 3 different builders (fly-builder-old-violet-4905, fly-builder-lingering-cloud-2611,
fly-builder-wispy-forest-8680). Two hit the original fdaa::3:5000 connection refused, one got an HTTP 500 (unable to upgrade to h2c, received 500).
cd backend && flyctl deploy --app retentionlab-api --remote-only
Celery Worker:
cd backend && flyctl deploy --config fly.worker.toml --app retentionlab-worker --remote-only
Frontend:
cd frontend && flyctl deploy --app retentionlab-app --remote-only --build-arg VITE_API_URL=https://retentionlab-api.fly.dev/api/v1
All three hit the same error — the Docker image builds and all layers push successfully, but the final registry push fails with:
ERROR: failed to push registry.fly.io/retentionlab-<app>:deployment-<id>:
failed to do request: Head "https://_api.internal:5000/v2/retentionlab-<app>/blobs/sha256:<hash>?ns=registry.fly.io":
dial tcp [fdaa::3]:5000: connect: connection refused
The builds are using Depot (Fly’s remote builder). The registry at _api.internal:5000 is refusing connections.
All three retries have now failed with the identical _api.internal:5000: connection refused error.