Deactivating an executor removes it from the control plane and stops it from polling. There are two flavors: disable the API key (a quick reversible cut) and delete the executor row (a permanent removal that requires the executor to host no nodes).Documentation Index
Fetch the complete documentation index at: https://docs.novacula.io/llms.txt
Use this file to discover all available pages before exploring further.
Prerequisites
- A role with deploy permission: owner or admin.
- For a row delete: no nodes assigned to the executor. Migrate or delete its nodes first.
Quick disable: revoke the API key
Use this when you want the executor to immediately stop reaching the control plane (compromised host, decommissioning a colo, …) without losing the row or its history.- Settings → API keys.
- Find the key the executor is using (cross-reference with the executor row’s “last used key” indicator).
- Click Disable.
- The executor’s next
syncExecutorreturns401. WithinHEARTBEAT_TIMEOUT_MS(30s), the executor row’s virtualstatusflips tooffline. - Hosted nodes remain in the control plane in their last reported state. They show as unreachable; lifecycle actions queue but won’t apply until the executor reconnects.
- Underlying processes on the infrastructure keep running — the agent / operator daemon doesn’t stop just because it can’t reach the control plane.
Permanent removal: delete the executor row
Use this when you’re decommissioning the host for good.Step 1 — drain hosted nodes
You cannot delete an executor that has nodes. For each node:- If you want the data preserved, migrate the spec to a different executor (delete + redeploy on the new target).
- If you don’t, stop the node, then delete it.
Step 2 — delete the row
From Executors → executor detail → Delete. The control plane:- Marks the row deleted (the API key tied to it is also disabled).
- Writes an
executor.deletedevent to the Events feed. - Releases the
(organizationId, name)slot — you can reuse the name for a new executor.
Step 3 — clean up the host
The control plane has no way to reach into your infrastructure, so the daemon and any leftover blockchain processes need to be removed manually.Agent
systemd units the agent created — see Agent CLI for the unit naming convention.
Operator
helm uninstall — kubectl delete pvc per claim if you want the storage back, or leave them for forensics.