Files

T

Jon Chery ed91d68fbf feat(P10): observability expansion — metrics + security headers (REQ-159)

New metrics:
- orca_jobs_running / orca_jobs_failed / orca_jobs_complete (gauges)
- orca_audit_chain_head (gauge, chain integrity)
- orca_drift_events_total, orca_ssh_errors_total (counters)
- orca_txn_apply_total, orca_txn_rollback_total (counters)
- orca_acl_denials_total (counter)

Security headers on metrics + healthz endpoints:
- X-Content-Type-Options: nosniff
- X-Frame-Options: DENY

New file: docs/metrics.md (Prometheus reference + scrape config)

---ci---
project: orca
phase: 10
milestone: v0.13
status: complete
requirements:
  covered: [159]
---/ci---

2026-08-10 13:44:04 +00:00

1.6 KiB

Raw Blame History

Orca Metrics Reference

Orca exposes Prometheus text-exposition metrics at /metrics on the metrics endpoint (default :9100, configurable via --addr).

Running the metrics endpoint

orca metrics --addr :9100

Prometheus scrape config

scrape_configs:
  - job_name: orca
    static_configs:
      - targets: ['localhost:9100']
    scrape_interval: 15s

Metric reference

Metric	Type	Description
`nodes_total`	Gauge	Total number of registered nodes
`allocs_total`	Gauge	Total number of job allocations
`orca_jobs_by_state{state}`	Gauge	Jobs grouped by status (running, complete, failed, etc.)
`orca_audit_chain_head`	Gauge	Audit chain integrity (1 = chain head verified, 0 = error)

Counter metrics (incremented by CLI operations)

The following counters are incremented during normal operations and are available when the metrics endpoint polls the DB:

Metric	Type	Description
`orca_drift_events_total`	Counter	Total drift events detected
`orca_ssh_errors_total`	Counter	Total SSH connection/exec errors
`orca_txn_apply_total`	Counter	Total transaction applies
`orca_txn_rollback_total`	Counter	Total transaction rollbacks
`orca_acl_denials_total`	Counter	Total ACL denials (enforce mode)

Security headers

The metrics endpoint sets the following security headers on all responses:

X-Content-Type-Options: nosniff
X-Frame-Options: DENY

Health check

The endpoint also exposes /healthz returning 200 ok for liveness probes.

1.6 KiB Raw Blame History