Skip to content
Tutorials Center
My learning

Topics

Lesson 33 of 37 ยท Patterns and Deployment

Logging, Monitoring and Deployment

Operate a Discord or Telegram bot in production: structured JSON logs, counters and a metrics endpoint, throttled error alerts, a CI deploy pipeline and a go-live checklist.

Once real people use your bot, three questions matter every day: is it up, is it healthy, and what happened when something went wrong? Logging answers the third, metrics answer the second, and a repeatable deployment keeps the first true. This lesson builds the small tools for all three and ends with a go-live checklist.

Structured logs

console.log("user 5 ran ping") is fine on a laptop. In production you will search and filter thousands of lines, so write logs as one JSON object per line (JSON Lines). Every line has a time, a level, a message and fields you can filter on, such as chatId or command.

Node.js A small structured logger with levels and child context

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with node:readline or process.stdin), it gets the next line from this box.

Notes on good logging:

  • A child logger carries the request context (update ID, chat, user) so every line from one event can be found together.
  • Log events, not chatter: start, stop, command handled (with duration), API errors, rate limits, retries.
  • Never log message content in a bot that handles private chats unless you must; user text is personal data. Log the command name and IDs, and say in your privacy notice what you keep and for how long.
  • Write to standard output and let the platform (systemd journal, Docker, a log shipper) collect it. Do not manage log files by hand.

Metrics: numbers over time

Logs tell stories; metrics tell trends. Counters (commands handled, errors, API calls, 429 responses) and a few gauges (queue length, last successful poll) are enough to begin with. Expose them in the simple text format that Prometheus and many hosted tools understand.

Node.js Counters and a /metrics endpoint

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with node:readline or process.stdin), it gets the next line from this box.

Keep label values to small fixed sets (command names, status codes). Never use a user ID or a chat ID as a label: a million users would create a million time series. Protect /metrics the same way as other internal endpoints, for example by listening on a private network or checking a token.

Useful alerts, in order of value: the bot has not polled or received an update for N minutes; error rate rises; 429 rate rises; restarts happen repeatedly; disk is nearly full (SQLite file or logs).

Telling a human about errors

Uncaught errors should reach you without flooding you. A common pattern is to send them to a private Telegram chat or Discord channel through a webhook, with throttling so a loop of identical errors produces one message, not a thousand.

Node.js Throttled error alerts

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with node:readline or process.stdin), it gets the next line from this box.

Deploying changes

A deployment should be boring and repeatable. A sensible minimal pipeline, shown read-only as a GitHub Actions file:

YAML
name: deploy
on:
  push:
    branches: [main]
jobs:
  test-and-deploy:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with: { node-version: 24 }
      - run: npm ci
      - run: node --test
      - run: docker build -t registry.example.com/mybot:${{ github.sha }} .
      - run: docker push registry.example.com/mybot:${{ github.sha }}
      # then: ssh to the server and run "docker compose pull && docker compose up -d"

Principles that keep deployments safe:

  • Tests run before anything ships. A red build never deploys.
  • Images are tagged with the commit, so rollback means running the previous tag.
  • Database migrations (SQLite storage) run at startup, are small and are backward compatible with the previous version of the code where possible.
  • Secrets live on the server or in the platform's secret store, never in the pipeline file or image.
  • After a deploy, watch the logs and the health endpoint (hosting lesson) for a few minutes.

Go-live checklist

  • Token and secrets come from the environment; none are in Git history.
  • Bot has only the permissions and intents it needs.
  • Webhook secret or interaction signature verification is on.
  • Rate limiting, retries with backoff and a send queue exist.
  • Data is on a volume with a tested backup and restore.
  • Graceful shutdown and a supervisor restart policy are configured.
  • Logs are structured, secrets are redacted, errors alert a human.
  • A privacy text explains what you store and how users can delete it, with a /delete_my_data or similar command.
  • Platform rules were read: Discord's developer policy and Telegram's bot guidelines.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. Why log JSON lines instead of free text?

    Structured fields make logs queryable by tools.
  2. 2. What is wrong with using a user ID as a metrics label?

    High-cardinality labels explode the number of series; use small fixed sets like command names.
  3. 3. What is the purpose of throttling error alerts?

    One alert plus a count of suppressed duplicates is readable; a thousand identical alerts are not.
  4. 4. How do you roll back quickly if images are tagged with the commit SHA?

    Immutable, commit-tagged images make rollback a one-line change.

Summary

  • Write structured JSON logs with child context and automatic redaction; log events, not private message text.
  • Expose counters and gauges at /metrics; keep label sets small.
  • Alert on silence and error spikes, with throttling.
  • Deploy through tests, commit-tagged images and compatible migrations, and keep a go-live checklist.

Next: Project: a reminder bot