Crusoe Webhooks

generated: '2026-08-04'
method: searched
source: https://docs.crusoecloud.com/notifications/overview
spec_type: none
asyncapi_published: false
asyncapi_note: >-
  Crusoe publishes no AsyncAPI document and no event-schema registry. The webhook catalog below is
  captured from the Notifications documentation. Nothing here is inferred — every event category,
  field, and error code is published by Crusoe.
surface:
  name: Crusoe Cloud Notifications
  docs: https://docs.crusoecloud.com/notifications/overview
  availability: all Crusoe Cloud users
  configured_in: Crusoe Cloud Console (bell icon → All Notifications → Configure Slack/Webhook Notifications)
  api_managed: false
  note: >-
    Endpoints are created in the Console UI only — there is no endpoint-management operation in the
    Crusoe Cloud API Gateway OpenAPI, so webhook subscriptions cannot be provisioned programmatically.
    Delivery is fronted by Svix (the docs instruct enterprise Slack admins to authorize the Svix app).
channels:
- name: console
  default: true
  description: All notifications appear in the Console notification sidebar and history.
- name: email
  default: true
  description: Sent automatically to all users in the organization for critical resource health alerts.
- name: slack
  default: false
  description: Routed to a Slack channel via an incoming webhook URL or the Connect to Slack flow.
- name: webhook
  default: false
  description: Generic HTTP POST endpoint for PagerDuty, Opsgenie, or custom automation.
  transport: HTTP POST
  content_type: application/json
  subscription: per-endpoint selection of event types
delivery:
  method: HTTP POST
  body: JSON
  schema_published: false
  signature_verification: not documented
  retry_policy: not documented
  example_payload:
    event_type: node_replacement_initiated
    cluster_id: cluster-abc123
    cluster_name: training-cluster-prod
    vm_id: vm-xyz789
    vm_name: worker-node-42
    error_code: XID 79
    description: GPU has fallen off the PCIe bus. AutoClusters has initiated node replacement.
    action: REPLACE_NODE
    timestamp: '2026-02-11T14:32:00Z'
    command_center_url: https://console.crusoecloud.com/orchestration/clusters/cluster-abc123/command-center
  example_source: https://docs.crusoecloud.com/notifications/overview
common_fields:
- {name: vm_id, description: Unique identifier of the affected node}
- {name: vm_name, description: Human-readable name of the affected node}
- {name: cluster_name, description: Cluster containing the affected node}
- {name: cluster_id, description: Unique identifier of the cluster}
- {name: event_type, description: Category of the event (replacement, maintenance, failure)}
- {name: error_code, description: Specific hardware error code, if applicable}
- {name: description, description: Brief explanation of what was detected}
- {name: action, description: Remediation action taken (automatic or manual)}
- {name: timestamp, description: When the event was detected}
- {name: command_center_url, description: Direct link to the relevant Console view}
event_categories:
- id: autoclusters-remediation
  name: AutoClusters remediation events
  requires: AutoClusters enabled on a Crusoe Managed Kubernetes cluster
  docs: https://docs.crusoecloud.com/orchestration/cmk/autoclusters
  events:
  - id: node_replacement_initiated
    description: A critical hardware failure was detected and remediation started; includes error code
      and affected node.
  - id: node_replacement_completed
    description: A faulty node was drained, removed, and replaced with a healthy node from the spare pool.
  - id: node_replacement_failed
    description: Remediation could not complete (e.g. no spare nodes available). Crusoe Cloud Support is
      automatically notified.
  - id: detection_only
    description: A hardware issue was detected but did not meet the threshold for automatic remediation;
      the error code is included for review.
  trigger_codes:
  - {code: GPUFellOffTheBus, description: GPU lost from PCIe bus}
  - {code: HCAFellOffTheBus, description: Host Channel Adapter (InfiniBand) lost}
  - {code: XID 48, description: Uncorrectable double-bit ECC memory error}
  - {code: XID 64, description: ECC page retirement or row remapper recording failure}
  - {code: XID 74, description: NVLink interconnect error}
  - {code: XID 79, description: GPU fell off the PCIe bus}
  - {code: XID 119, description: GSP not responding to driver RPC requests}
  - {code: XID 120, description: Driver failed to recover from GSP timeout}
  detection_only_codes:
  - {code: XID 76, description: Internal micro-controller breakpoint}
  - {code: XID 94, description: Contained ECC error}
  - {code: XID 95, description: Uncontained ECC error}
  - {code: XID 137, description: Unexpected completion}
  - {code: XID 140, description: Unrecovered ECC error}
  - {code: XID 143, description: GPU initialization failure}
- id: gpu-xid-standalone
  name: GPU XID alerts for standalone VMs
  remediation: informational only, no automatic remediation
  trigger_codes:
  - {code: XID 48, description: Uncorrectable double-bit ECC memory error}
  - {code: XID 64, description: ECC page retirement or row remapper recording failure}
  - {code: XID 74, description: NVLink interconnect error}
  - {code: XID 79, description: GPU fell off the PCIe bus}
  - {code: XID 95, description: Uncontained ECC error}
  - {code: XID 119, description: GSP not responding to driver RPC requests}
  - {code: XID 120, description: Driver failed to recover from GSP timeout}
  - {code: XID 140, description: Unrecovered ECC error}
  - {code: XID 143, description: GPU initialization failure}
- id: critical-failure
  name: Critical failure events
  description: >-
    Critical hardware failures requiring manual intervention on nodes not covered by AutoClusters
    (non-GPU nodes, or clusters without AutoClusters enabled).
- id: budget-alerts
  name: Budget alerts
  description: Fired when spend reaches a predefined threshold.
  docs: https://docs.crusoecloud.com/usage-billing/budget-alerts
planned:
- Inference service notifications
- Organization-level alerts
gaps:
- No AsyncAPI or JSON Schema document for any event payload.
- No published webhook signature/verification scheme.
- No published retry or delivery-guarantee policy.
- Webhook endpoints cannot be created, listed, or deleted through the REST API.
x-evidence:
  fetched: '2026-08-04'
  url: https://docs.crusoecloud.com/notifications/overview
  http_status: 200