Files
mesh-catalog/modules/mesh-watcher
mesh-tools 652b00b25b
mesh/merge-gate checking again: the gate refused `unknown` because the controller judged with a node-engine validator pinned before mesh-host #117; mesh-co…
mesh/repo-check checking again: the gate refused `unknown` because the controller judged with a node-engine validator pinned before mesh-host #117; mesh-co…
mesh/delivery-group group feat/owned-paths checking: 0 of 2 member(s) ready
mesh/delivery superseded: a newer head of the same pull request
Say on every directory how it treats files it does not know, and check it (issue 1024, ADR 0422)
Every directory and archive resource carries `unknown`: move-out for a home folder whose program writes nothing there (the node-engine, delivered in mesh-host #117, moves the rest to ~/.local/state/mesh-moved/<module>/ with a list), module for ~/.ssh and ~/.claude, program elsewhere; `knows` keeps what a program or the operator relies on (powerlevel10k's compiled files, i3's scripts and the asus/g14 drop-ins, dunst's slack rule, betterbird's records, node-env's forge credential, issue 1033). Two shared folders are no longer declared. Each module's README says it. checks/owned holds every manifest to it; checks/sshdir stays while main's merge check runs it (issue 1035). Tests that pin a manifest follow the new key only. Reviewed in three rounds.
2026-10-17 16:23:03 +00:00
..

mesh-watcher

The watcher's watcher (novox/hq to-be 45 §5, signal S10, ADR 0227 rule 6): what tells the operator when the parts that would tell them are what failed.

  • Assigned to one machine that is not the control node. It reads where the controller and the bus run (mesh-controller.seats) every ten minutes; on the same machine its status says MISPLACED.
  • Watches two signals.
    • The self-check: the controller's doctor heartbeat, mesh-controller.doctor-heartbeat. Bound: twice the interval the heartbeat says doctor runs at (five minutes when it says none). Heard by the time the heartbeat says it was made, so a backlog delivered after a restart is not a sign of life. The heartbeat is read in one place, cmd/mesh-watcher/heartbeat.go, which states every assumption it makes about its shape.
    • The bus: a round trip through the bus server — its own watcher_ping on its own machine — every minute. Bound: three minutes.
  • Silent past its bound: it sends to Telegram directly over HTTPS, not through the bus, once; once more an hour later if still silent; and again when the signal returns. Before a signal is first heard it counts from the watcher's start: a heartbeat that never comes is what this is for.
  • Its own health is visible: watcher_status leads with whether it can tell the operator anything at all (BLIND without a token or chat id, or while Telegram fails), whether it is misplaced, and whether heartbeats reach it. A message it could not send is owed and tried every minute.

What the operator gives

  1. The bot token, as this module's own secret: secret accept <machine> mesh-watcher telegram-token, then push that machine. The same bot as the operator-channel's is fine; it is one more machine holding it (ADR 0227, accepted for this one case).
  2. The chat id, as a setting: settings for mesh-watcher with {"telegram-chat-id": "<id>"}.

Tools

tool does
watcher_status its own health, then each signal: last heard, bound, silent, owed
watcher_last_heard when each signal was last heard, and what it sent lately
watcher_test a test message to Telegram now, directly
watcher_ping the bus round trip's other end

Files it does not know

Each directory or archive this module declares says how it treats a file it did not write (ADR 0422, ADR 0416).

  • ${dir:state}: everything in it is the program's; the mesh moves nothing out (ADR 0422 §2).