Skip to content

Latest commit

 

History

History
88 lines (66 loc) · 2.95 KB

configure-health-checks.md

File metadata and controls

88 lines (66 loc) · 2.95 KB
Raw

Health Checks

Introduction

Styx automatically monitors the connectivity and health of configured backend services. This feature is called "health checking", and can be enabled or disabled by including a healthCheck attribute in a backend service configuration block.

Health checking is configured for a specific backend service, using the uri attribute to indicate the URI of the health check endpoint. Styx polls this endpoint, for every backend service origin, at configured intervals.

When an origin fails to respond to the poll from Styx for unhealthyThreshold times, Styx will mark the origin as INACTIVE and remove it from the load balancer rotation. Conversely, when an INACTIVE origin starts responding to the polls, it must respond at least healthyThreshold times before Styx moves it back into the ACTIVE state and puts it back into the load balancer rotation.

Origin State Machine

An origin can be in one of the three possible states:

  • DISABLED - This is an administrative state. A system administrator has removed an origin from load balancer rotation.

  • ACTIVE - Origin is administratively enabled. It is responding to styx health checks, and considered active. Styx will load balance traffic to the origin.

  • INACTIVE - Origin is administratively enabled but it is not responding to the health checks and therefore considered inactive. Styx will not load balance traffic to the origin.

A note about enabling a DISABLED origin: to prevent Styx from sending traffic to a potentially broken origin, the origin is initially enabled in an INACTIVE state. It gets activated only after health checks have confirmed the origin is healthy.

However, if health checks for the backend service are disabled, the origin will be activated straight away.

    DISABLED <----------- ACTIVE
       ^                    ^
       |                    |
       |                    |
       |                    v
       `--------------> INACTIVE

Configuration

To enable health checking, add the healthCheck configuration block into BackendService.

healthCheck:
  uri: "/version.txt"
  intervalMillis: 10000
  healthyThreshold: 2
  unhealthyThreshold: 2
  • uri
    • An URI endpoint used for the health check poll.
  • intervalMillis
    • Time between two consecutive health check polls.
  • healthyThreshold
    • Number of consecutive successful health checks before an INACTIVE origin is activated (goes into ACTIVE state).
  • unhealthyThreshold
    • Number of consecutive unsuccessful health checks before an ACTIVE origin is deactivated (goes into INACTIVE state).

Metrics

A meter of failed health check attempts per backend service:

origins.<BACKEND-ID>.healthcheck.failure.count
origins.<BACKEND-ID>.healthcheck.failure.m1_rate
origins.<BACKEND-ID>.healthcheck.failure.m5_rate
origins.<BACKEND-ID>.healthcheck.failure.m15_rate
origins.<BACKEND-ID>.healthcheck.failure.mean_rate