One API key per feeder + feeder-scoped ingest URLs (migrate off shared keys)

Open
  • Is something wrong?
#438 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
5/5
Estimated time
Over a week
Newbie friendliness
12/100
Issue type
Feature
Clarity
Mostly clear
Activity status
Active
Tech stack
django, prometheus, python
Domain
api, authentication, backend

Research direction

Start with the auth path: NodeAuthorizationPermission and the NodeAuth link between NodeAPIKey and ManagedNode, then common/meshcore_feeder_auth.py for the MeshCore equivalent. The issue's open points need a maintainer decision first, including the per-protocol key rule and the prod query for keys linked to several feeders. Done means the acceptance criteria are met: feeder-scoped endpoints reject shared keys, old v1/v3 URLs are unchanged, and Prometheus shows usage of old and new URLs.

Written by the indexing model from the issue text.

Description

enhancement meshtastic-2.8 permissions

Part of #437. This ticket should land before the feeder re-key work (ID-4) and makes that work much smaller.

Problem

Today an API key can be shared between several feeders owned by the same person (NodeAuth is many-to-many between NodeAPIKey and ManagedNode). Because of that, the feeder's node number in the URL (/api/packets/{node_id}/...) is what chooses which feeder is posting, and NodeAuthorizationPermission authorises on the pair (key, node number).

From 2.8, a node's number is crc32(public_key), so it changes on upgrade. An upgraded feeder then posts under a number its key isn't linked to, gets a 403, and ingest stops silently.

MeshCore has the same sharing built in: common/meshcore_feeder_auth.py uses the X-MeshCore-Feeder-Pubkey header to tell a user's feeders apart.

Proposal: one feeder, one API key

When a key belongs to exactly one feeder, the key on its own identifies the ManagedNode. The node number the bot reports is then just information we record, not something we authorise on. If the number changes, the feeder simply reports a new one:

  • if crc32(public_key) equals the new number, update meshtastic_node_id automatically, keep the old number as an alias and notify the owner;
  • if it doesn't match, flag it for review rather than reject it. The likely cause is a bot config copied onto another radio.

Keys can still be rotated.

Migration

  1. API: add feeder-scoped endpoints with no node number in the path, for example /api/v4/feeder/ingest/, /api/v4/feeder/nodes/ and /api/v4/feeder/bot-version/. Exact paths to be decided; MeshCore gets the equivalent. These endpoints require a key linked to exactly one feeder. A shared key gets a clear 4xx telling the owner to mint a per-feeder key.
  2. Bot: post to the new URLs. Ship this in the same bot release as feeder identity reporting (ID-5).
  3. Owners upgrade their bot over time. The old v1/v3 URLs keep working unchanged, shared keys included. No new shared links can be created (UI/admin), though existing ones keep working on the old URLs.
  4. Track usage of the old and new URLs in Prometheus:
    • Request counts per URL version. django-prometheus per-view metrics may already cover this; check the views have distinct names.
    • Which feeders are still on the old URLs, so we can chase their owners. Use either a low-cardinality counter labelled by feeder (the fleet is small) or a last_used_legacy_url_at field on ManagedNode. Pick one.
  5. Retire the old URLs once legacy usage reaches zero, or stays at a known handful we've contacted. This joins up with #319 (retire v1 ingest URLs).
  6. Once the old URLs are gone, enforce one key per feeder in the schema: a unique constraint on NodeAuth.api_key, or move the key onto ManagedNode.

Open points

  • A bot running both a Meshtastic and a MeshCore radio on one key: should the rule be one key per feeder per protocol, or one bot record owning both radios? Check prod for keys linked across protocols.
  • Before step 6, find out how many prod keys are linked to more than one feeder:
    NodeAuth.objects.values("api_key").annotate(n=Count("node")).filter(n__gt=1)
  • Dual-API (prod plus pre-prod) is unaffected, because each environment has its own key.

Acceptance criteria

  • New feeder-scoped endpoints authenticate by key alone and reject shared keys with an actionable error.
  • A feeder whose node number changes keeps ingesting on the new endpoints. Its number is updated, with the CRC check, an alias for the old number, and an owner notification. If ID-4 lands separately, at minimum there is no 403.
  • Old v1/v3 endpoints behave exactly as they do today.
  • New shared key links can no longer be created.
  • Prometheus shows request volume for the old and new URLs, plus a way to list the feeders still on the old URLs.
  • Bot change tracked in a linked meshflow-bot issue.
  • Docs updated: API.md, API_KEYS.md and the ingestion docs.
Dominant language
Python
Stars
1
Forks
0
Avg merge
1d 20h
Merged PRs (30d)
5

Getting set up

  • Ships a Dockerfile or Docker Compose file
  • Has a pull request template
  • No contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from pskillen/meshflow-api

All issues in pskillen/meshflow-api

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.