Documentation
¶
Index ¶
- Constants
- Variables
- type MetricServer
- type Probe
- type ProbeRegistry
- func (p *ProbeRegistry) Liveness(ctx context.Context) error
- func (p *ProbeRegistry) Readiness() error
- func (p *ProbeRegistry) Refresh(ctx context.Context)
- func (p *ProbeRegistry) RegisterLivenessProbe(key string, probe Probe)
- func (p *ProbeRegistry) RegisterReadiness(key string) *Readiness
- func (p *ProbeRegistry) RegisterReadinessProbe(key string, probe Probe)
- func (p *ProbeRegistry) Run(ctx context.Context) error
- func (p *ProbeRegistry) WithdrawReadiness()
- type Readiness
Constants ¶
const (
ModuleName = "observability"
)
Variables ¶
var ErrNotEvaluated = fmt.Errorf("%w: probes have not run", ErrNotReady)
ErrNotEvaluated is what Readiness returns before the probes have run for the first time. Seeing it for longer than that means nothing is running the ProbeRegistry.
var ErrNotReady = errors.New("not ready")
ErrNotReady is what readiness probes return for expected states, e.g. before startup has finished or once shutdown has started. The readiness endpoint logs these rather than reporting them as processing errors, which would send one event per health check to Sentry on every deploy.
var ErrShuttingDown = fmt.Errorf("%w: shutting down", ErrNotReady)
ErrShuttingDown is what Readiness returns once readiness has been withdrawn.
Functions ¶
This section is empty.
Types ¶
type MetricServer ¶
type MetricServer struct {
Log *slog.Logger
API modshared.API
// Name is the name of the application.
Name string
Listener net.Listener
PrometheusURLPath string
LivenessProbeURLPath string
ReadinessProbeURLPath string
Gatherer prometheus.Gatherer
Registerer prometheus.Registerer
ProbeRegistry *ProbeRegistry
}
func (*MetricServer) ConstructHandler ¶
func (s *MetricServer) ConstructHandler() http.Handler
type ProbeRegistry ¶
type ProbeRegistry struct {
// contains filtered or unexported fields
}
func NewProbeRegistry ¶
func NewProbeRegistry() *ProbeRegistry
func (*ProbeRegistry) Readiness ¶
func (p *ProbeRegistry) Readiness() error
Readiness returns the verdict of the last probe run. Reading it is a memory access: readiness is served to whoever can reach the health endpoints, so a caller must not be able to drive a probe run - and a Redis round trip - per request. Every caller gets the same verdict, so all health endpoints agree.
Run() has to be running for the verdict to be updated.
func (*ProbeRegistry) Refresh ¶ added in v19.4.0
func (p *ProbeRegistry) Refresh(ctx context.Context)
Refresh runs the readiness probes and updates the verdict. Runs are serialized, so concurrent calls cannot publish out of order.
func (*ProbeRegistry) RegisterLivenessProbe ¶
func (p *ProbeRegistry) RegisterLivenessProbe(key string, probe Probe)
func (*ProbeRegistry) RegisterReadiness ¶ added in v19.4.0
func (p *ProbeRegistry) RegisterReadiness(key string) *Readiness
RegisterReadiness registers a readiness probe for a component that reports its own readiness. The component is not ready until it says so, and can say it is not ready again, e.g. when it starts shutting down.
func (*ProbeRegistry) RegisterReadinessProbe ¶
func (p *ProbeRegistry) RegisterReadinessProbe(key string, probe Probe)
func (*ProbeRegistry) Run ¶ added in v19.4.0
func (p *ProbeRegistry) Run(ctx context.Context) error
Run refreshes the readiness verdict until ctx is done.
func (*ProbeRegistry) WithdrawReadiness ¶ added in v19.4.0
func (p *ProbeRegistry) WithdrawReadiness()
WithdrawReadiness makes Readiness report that this instance is shutting down. It doesn't run the probes, so it takes effect immediately rather than at the end of a probe run, and it can't be undone: an instance that has begun shutting down doesn't become ready again.
type Readiness ¶ added in v19.4.0
type Readiness struct {
// contains filtered or unexported fields
}
Readiness is the readiness of one component, reported by the probe registered alongside it.
func (*Readiness) SetNotReady ¶ added in v19.4.0
func (r *Readiness) SetNotReady()
Source Files
Directories