Production Guides & Recipes

Container Resource Checks in Kubernetes / Docker

When running inside containers, Node’s os.totalmem() reports the host machine’s total RAM rather than the container’s cgroup memory limit. @nexload-sdk/healthcheck-node automatically inspects Linux cgroups to prevent silent out-of-memory (OOM) kills:

import { containerResourceCheck } from "@nexload-sdk/healthcheck-node";

const check = containerResourceCheck({
  memoryUsageWarningPercent: 80,  // "degraded" when above 80%
  memoryUsageCriticalPercent: 95, // "unhealthy" when above 95%
  cpuThrottleWarningPercent: 20,  // "degraded" when throttled >20%
});

---

TCP & DNS Dependency Probes

Verify external database and downstream microservice reachability during readiness probes:

import { dnsCheck, tcpCheck } from "@nexload-sdk/healthcheck-node";

// 1. Verify TCP port connectivity (e.g. Postgres or Redis)
const redisTcp = tcpCheck({
  name: "redis.connection",
  host: process.env.REDIS_HOST ?? "localhost",
  port: 6379,
  timeoutMs: 2000,
  scopes: ["readiness"],
});

// 2. Verify external DNS resolution
const stripeDns = dnsCheck({
  name: "dns.stripe",
  hostname: "api.stripe.com",
  timeoutMs: 1500,
  scopes: ["readiness"],
});

---

Troubleshooting & Common Pitfalls

1. cgroup Resource Check Reports “Degraded” on Non-Linux Hosts

On macOS or Windows development machines, Linux cgroup controllers do not exist. @nexload-sdk/healthcheck-node safely falls back to host OS metrics and reports a low-confidence degraded flag in dev environments. In production Linux containers, full cgroup v1/v2 metrics are parsed automatically.

2. TCP Probe Leaks Sockets

tcpCheck immediately destroys the created socket once the TCP handshake completes or times out. Ensure external firewalls allow transient health check connection attempts.