Error: getaddrinfo EAI_AGAIN (temporary DNS failure)

getaddrinfo EAI_AGAIN host.example.com means the system resolver could not get an answer right now: EAI_AGAIN is getaddrinfo's "temporary failure in name resolution". Unlike ENOTFOUND, which says the name does not exist, EAI_AGAIN says the lookup itself failed (no reply from the DNS server, a timeout, SERVFAIL). Check network and DNS configuration in the environment where the code runs (container, CI job, VPN), and retry transient failures.

How it appears

Node.js prints the message Error: getaddrinfo EAI_AGAIN registry.npmjs.org followed by the stack, and the error object carries code: 'EAI_AGAIN', syscall: 'getaddrinfo' and the hostname that failed. Check err.code in code, not the message text.

The same failure from other programs on Linux: Temporary failure in name resolution (curl, ping, Python's socket.gaierror: [Errno -3]), and npm error code EAI_AGAIN from npm.

Diagnose

Run these in the same place the error happens (inside the container: docker exec -it NAME sh):

getent hosts registry.npmjs.org        # uses the same resolver path as getaddrinfo
cat /etc/resolv.conf                   # which DNS servers are configured
nslookup registry.npmjs.org 1.1.1.1    # does a public resolver answer?
dig +short registry.npmjs.org          # if dig is installed

If a public resolver answers but the configured one does not, the configured DNS server is the problem. If nothing answers, the network itself (route, firewall, proxy) is.

Fixes by environment

Retry transient failures

async function withRetry(fn, attempts = 3) {
  for (let i = 1; ; i++) {
    try { return await fn(); }
    catch (e) {
      if (e.code !== "EAI_AGAIN" || i >= attempts) throw e;
      await new Promise((r) => setTimeout(r, 200 * 2 ** i));
    }
  }
}

Retrying ENOTFOUND is pointless (the name does not exist); retrying EAI_AGAIN a few times with backoff often succeeds after a network change.