Log to a rotating file in production

You want production logs on disk: one file per day, capped so a single file never grows without bound, and guaranteed to flush the last entries when the container stops. That’s three options on FileLog plus one autoFlushOn.

The setup

import { log, ConsoleLog, FileLog } from "@warlock.js/logger";

const isProduction = process.env.NODE_ENV === "production";

log.configure({
  channels: isProduction
    ? [
        new FileLog({
          storagePath: process.cwd() + "/storage/logs",
          chunk: "daily", // one file per calendar day → 2024-03-15.log
          rotate: true, // roll the file over once it gets too big
          maxFileSize: 10 * 1024 * 1024, // 10 MB per file before rotation
          retention: { maxFiles: 30, maxAgeDays: 30 }, // prune old days and archives
        }),
      ]
    : [new ConsoleLog()],
  // Drain the in-memory buffer before the process exits.
  autoFlushOn: ["SIGINT", "SIGTERM", "beforeExit"],
});

Import it once, as early as possible:

import "./logger"; // side-effect: configures the singleton

import { log } from "@warlock.js/logger";

log.info("app", "boot", "Server listening on :3000");

What each piece does

chunk: "daily" names the file by ISO date, so you get 2024-03-15.log, 2024-03-16.log, … (sorted by name = sorted by day) instead of one ever-growing app.log. Use "hourly" for very high volume, "single" (the default) for a single file, or fileName: "app-{YYYY}-{MM}-{DD}" for your own pattern.
rotate: true (the default) checks the active file’s size before each write and, once it passes maxFileSize, renames it aside (2024-03-15-2024-03-15-1710494523000.log) and starts a fresh one.
retention removes this channel’s old files — on boot and whenever a new file starts — keeping the 30 newest and none older than 30 days. Only files matching the channel’s name pattern are touched; failures go to onChannelError, never to the call site.
maxFileSize is in bytes — 10 * 1024 * 1024 is 10 MB, which is also the default.
autoFlushOn is the important one. FileLog buffers entries in memory (it flushes at 100 entries or every 5 seconds), so a process that exits without draining loses whatever is still buffered. Wiring SIGINT / SIGTERM / beforeExit makes the logger flush on Ctrl+C, docker stop, and natural exit. See Shutdown & flushing.

Keeping errors in their own file

A common refinement: everything goes to the daily file, and errors and warnings get their own dedicated file so on-call doesn’t have to grep the firehose.

log.setChannels([
  new FileLog({ storagePath: "./storage/logs", chunk: "daily" }),
  new FileLog({
    storagePath: "./storage/logs/errors", // ← its own directory
    levels: ["error", "warn"], // only these reach this file
    chunk: "daily",
  }),
]);

The levels whitelist is per channel — the first file still records everything. See Channels Overview for the filtering model.