Health Checks
Health checks let load balancers, orchestrators, and monitoring systems know if your server is ready to handle traffic. Prism’s PrismHealthMonitor runs checks on demand and serves results through a /health endpoint.
Quick Setup
let health = PrismHealthMonitor()
await server.use(PrismHealthMiddleware(monitor: health))
// GET /health → {"status": "healthy", "checks": []}
Adding Checks
let health = PrismHealthMonitor()
// Simple status check
await health.register("memory") {
let info = ProcessInfo.processInfo
return info.physicalMemory > 0 ? .healthy : .unhealthy
}
// Detailed check with messages
await health.register(PrismHealthCheck(name: "database") {
do {
try await db.execute("SELECT 1")
return PrismHealthCheckResult(name: "database", status: .healthy, message: "Connected")
} catch {
return PrismHealthCheckResult(name: "database", status: .unhealthy, message: error.localizedDescription)
}
})
await health.register(PrismHealthCheck(name: "disk") {
let free = freeDiskSpace()
if free < 100_000_000 {
return PrismHealthCheckResult(name: "disk", status: .unhealthy, message: "Less than 100MB free")
} else if free < 1_000_000_000 {
return PrismHealthCheckResult(name: "disk", status: .degraded, message: "Less than 1GB free")
}
return PrismHealthCheckResult(name: "disk", status: .healthy)
})
await server.use(PrismHealthMiddleware(monitor: health))
Health Status Levels
| Status | HTTP Code | Meaning |
|---|
.healthy | 200 | Everything is operational |
.degraded | 503 | Running but with issues — may need attention |
.unhealthy | 503 | Not ready to handle traffic |
.unhealthy, the report is unhealthy. If any is .degraded (and none unhealthy), the report is degraded.
{
"status": "degraded",
"checks": [
{"name": "database", "status": "healthy", "message": "Connected"},
{"name": "disk", "status": "degraded", "message": "Less than 1GB free"},
{"name": "memory", "status": "healthy"}
]
}
Custom Endpoint Path
await server.use(PrismHealthMiddleware(monitor: health, path: "/api/health"))
// Now at GET /api/health instead of /health
Programmatic Health Checks
Run checks without the middleware:
let report = await health.checkHealth()
if report.status == .unhealthy {
print("Server is unhealthy!")
for check in report.checks where check.status == .unhealthy {
print(" Failed: \(check.name) — \(check.message ?? "unknown")")
}
}
Kubernetes uses two kinds of probes: liveness (is the process alive?) and readiness (can it handle traffic?). Use /health for readiness probes. For liveness, a simple 200-OK endpoint that doesn’t check dependencies is usually enough.
Each check’s execution time is measured automatically. This helps identify slow dependencies — a database check that takes 5 seconds is a red flag even if it returns healthy.
