Lifecycle Policies¶

Lifecycle policies control how Kubernetes checks the health of your service. These are timing presets for startup, readiness, and liveness probes.

What probes check¶

Probe	When	Why
Startup	On container start	Waits for the service to initialize
Readiness	Continuously	Determines if the service is ready to accept traffic
Liveness	Continuously	Restarts the container if the service is stuck

All probes use an HTTP GET request to healthPath and port from .hive.yml.

Built-in policies¶

`standard` (default)¶

Suitable for most services. Startup within ~60 seconds, checks every 10 seconds.

# .hive.yml
lifecycle: standard    # can be omitted, this is the default

`relaxed`¶

For heavy services with long startup times (ML models, large .NET applications, services with DB migrations).

lifecycle: relaxed

`aggressive`¶

For lightweight services that start instantly. Fast problem detection.

lifecycle: aggressive

Timing comparison¶

Startup Probe¶

Parameter	standard	relaxed	aggressive
`periodSeconds`	2	10	1
`failureThreshold`	30	60	30
Max startup time	60s	600s	30s

Readiness Probe¶

Parameter	standard	relaxed	aggressive
`periodSeconds`	10	30	5
`timeoutSeconds`	5	30	3
`successThreshold`	1	1	1
`failureThreshold`	3	5	1

Liveness Probe¶

Parameter	standard	relaxed	aggressive
`periodSeconds`	10	30	5
`timeoutSeconds`	5	30	3
`failureThreshold`	3	5	3

How to choose¶

graph TD
    A[Your service] --> B{Starts quickly?<br/>< 5 seconds}
    B -->|Yes| C{Need fast<br/>failover?}
    B -->|No| D{Takes longer than<br/>60 seconds?}
    C -->|Yes| E[aggressive]
    C -->|No| F[standard]
    D -->|Yes| G[relaxed]
    D -->|No| F

Your case	Policy
Regular web service (Go, Node.js, Python)	`standard`
Fast stateless service	`aggressive`
.NET with long initialization	`relaxed`
ML service, loading models	`relaxed`
Service with DB migrations on startup	`relaxed`

Health endpoint¶

The policy defines only timings. Path and port are service properties:

# .hive.yml
port: 8080
healthPath: /healthz     # default: /
lifecycle: standard

The endpoint should:

Return HTTP 2xx (200, 204, ...)
Respond quickly (within timeoutSeconds)
Be lightweight (no heavy DB queries)

Simple health endpoint

No need to check all dependencies. Just return 200 OK if the service is running and can process requests.

Custom policies¶

For non-standard cases, contact the platform team — they can add a custom preset to the Helm chart.