I have an orphaned machine in my app rag-prd that shows up in the GraphQL API but returns 404 from every Machines API endpoint. The
dashboard overview counts it (showing 6 machines), but the machines table only shows 5.
Orphaned machine:
- Machine ID: d8d1670b2e1648
- Name: green-sunset-4075
- State: failed
- Region: ord
- Created: 2025-11-04T23:30:58Z
- Last updated: 2026-01-11T07:53:04Z
- App: rag-prd
How I confirmed it exists:
The machine appears in the GraphQL API:
query {
app(name: “rag-prd”) {
machines {
totalCount
nodes { id name state }
}
}
}
This returns 6 non-destroyed machines, including d8d1670b2e1648 in failed state.
However, flyctl machines list -a rag-prd returns only 5 machines, and the Machines REST API with include_deleted=true also returns only
5:
curl -H “Authorization: Bearer $TOKEN”
“https://api.machines.dev/v1/apps/rag-prd/machines?include_deleted=true”
Returns 5 machines — the failed one is absent
Everything I’ve tried (all failed):
┌────────────────────────────────────────────────────────────────────────────────────┬────────────────────────────────────────────────┐
│ Approach │ Result │
├────────────────────────────────────────────────────────────────────────────────────┼────────────────────────────────────────────────┤
│ flyctl machine destroy d8d1670b2e1648 --force -a rag-prd │ “machine not found” │
├────────────────────────────────────────────────────────────────────────────────────┼────────────────────────────────────────────────┤
│ DELETE https://api.machines.dev/v1/apps/rag-prd/machines/d8d1670b2e1648?force=true │ 404 │
├────────────────────────────────────────────────────────────────────────────────────┼────────────────────────────────────────────────┤
│ Same with deploy token (fly tokens deploy) │ 404 │
├────────────────────────────────────────────────────────────────────────────────────┼────────────────────────────────────────────────┤
│ flyctl machine leases clear d8d1670b2e1648 -a rag-prd │ “machine not found” │
├────────────────────────────────────────────────────────────────────────────────────┼────────────────────────────────────────────────┤
│ REST API /stop, /signal, /cordon on the machine │ All 404 │
├────────────────────────────────────────────────────────────────────────────────────┼────────────────────────────────────────────────┤
│ Internal API _api.internal:4280 from a running machine via fly machine exec │ 404 │
├────────────────────────────────────────────────────────────────────────────────────┼────────────────────────────────────────────────┤
│ GraphQL removeMachine mutation │ “no longer supported” │
├────────────────────────────────────────────────────────────────────────────────────┼────────────────────────────────────────────────┤
│ GraphQL killMachine mutation │ “no longer supported” │
├────────────────────────────────────────────────────────────────────────────────────┼────────────────────────────────────────────────┤
│ fly scale count app=1 then fly scale count app=5 │ Only affects managed machines, orphan persists │
├────────────────────────────────────────────────────────────────────────────────────┼────────────────────────────────────────────────┤
│ fly apps restart rag-prd --force-stop │ Only targets running machines │
└────────────────────────────────────────────────────────────────────────────────────┴────────────────────────────────────────────────┘
Impact: The machine is not consuming compute or blocking deployments — it’s a cosmetic dashboard count issue (shows 6 instead of 5).
However, I’d like to clean it up.
Request: Could a Fly.io engineer remove machine d8d1670b2e1648 from the backend for app rag-prd? This appears to be the same known issue
described in several prior threads where machine metadata persists in the central database but the Machines API cannot reach it.
Thanks!