Healthchecks and SLAs

Healthchecks

A healthcheck is a second command that runs only after the main command exits with code 0.

healthcheck:
  command: "./scripts/healthcheck.sh"
  timeout: "30s"
  on_fail: mark_failed

on_fail behavior

Value	Meaning
mark_failed	Fail the run and apply retry policy
warn_only	Keep the run successful but mark hc_status=warn

Important runtime rule

If the main command fails, Husky does not run the healthcheck.

Healthcheck logs

Healthcheck output is stored separately in the log stream and is hidden from normal log output unless requested.

husky logs <job>
husky logs <job> --include-healthcheck

SLA budgets

An SLA is a soft duration budget for observability.

sla: "5s"
timeout: "30s"

If a job exceeds its SLA while still running:

• Husky marks the run as SLA-breached
• on_sla_breach notification can fire
• the job keeps running

Constraint

When both are set:

• sla must be less than timeout

Where SLA state appears

• husky history
• /api/runs/<id>
• dashboard run state and run history

Example

jobs:
  slow_job_with_sla:
    description: "Long-running job with early warning"
    frequency: manual
    command: "./scripts/slow.sh"
    timeout: "30s"
    sla: "5s"
    notify:
      on_sla_breach:
        channel: webhook:http://127.0.0.1:9999/sla
        message: "{{ job.name }} is still running after {{ run.elapsed }}"