mesh-sdk/go/delivery/nodehealth.go

package delivery

import (
	"bytes"
	"encoding/json"
	"errors"
	"fmt"
	"time"
)

// **A node's own health report on the bus** (issue 774, the operator's alert table of 2026-10-29): one key per node,
// `<node>`, of the node-engine's state `health` (NodeHealthBucket), written by that node's engine alone and only when
// what it says changed. The watchman (seat mesh-watchdogs) reads it for six alerts the controller raised from the
// passing health message: machine-network (Network), network-rewritten (the resolver file's part, Writer and Owner),
// network-unreachable (Peers, read across every node's report), machine-units (Units), relogin-needed (Relogin) and
// directory-used-as-found (Found). Moved here from mesh-host internal/link (65fbc1e6) unchanged; the engine builds a
// report from its own health statement there (NodeHealthOf).

// NodeHealthState and NodeHealthBucket are the node-engine's state health, as a reader names it and as the bus does.
const (
	NodeHealthState  = "health"
	NodeHealthBucket = "mesh-host_" + NodeHealthState
	NodeHealthRead   = "mesh-host." + NodeHealthState
)

// NodeHealthVersion is the version of NodeHealth's layout.
const NodeHealthVersion = 1

// NodeHealth is one node's health report, as its engine last found it.
type NodeHealth struct {
	V    int    `json:"v"`
	Node string `json:"node"`
	// Network is the machine's networking, its parts without their streaks; nil where the engine does not judge it yet.
	Network *NetworkHealth `json:"network,omitempty"`
	// Peers are the mesh tunnel's peers on the last look that read them, sorted by key.
	Peers []PeerSeen `json:"peers,omitempty"`
	// Units is the machine's service managers: each failed unit no module places.
	Units *UnitsHealth `json:"units,omitempty"`
	// Relogin are the accounts whose running session needs a new login for a group it was given.
	Relogin []NodeAccount `json:"relogin,omitempty"`
	// Found are the directories used as found.
	Found []NodeDirectory `json:"found,omitempty"`
	// At is when the engine wrote this, on a change: never a reason to write.
	At time.Time `json:"at"`
}

// NetworkHealth is the machine's networking (ADR 0241): the worst of its parts, since when, and each part.
type NetworkHealth struct {
	State string        `json:"state"`
	Since time.Time     `json:"since"`
	Parts []NetworkPart `json:"parts"`
}

// NetworkPart is one part of a machine's networking, as its engine judged it on its second look.
type NetworkPart struct {
	// Part is resolv-conf, names, tunnel, bus or route.
	Part  string `json:"part"`
	State string `json:"state"`
	// Reason is why it is not healthy, in words with no address, path or domain; Said the detail.
	Reason string `json:"reason,omitempty"`
	Said   string `json:"said,omitempty"`
	// Writer is the program that rewrote the resolver file, when it can be named; Owner the module whose file it is.
	Writer string `json:"writer,omitempty"`
	Owner  string `json:"owner,omitempty"`
	// Toward is what the failure points at: "hub", or each resolver's address that failed.
	Toward []string `json:"toward,omitempty"`
	// Targets are the mesh machines the failure points at, by node name, as the controller holds a part at the machine
	// it points at: the hub's machine for "hub", and the machine whose tunnel address is each resolver's. Empty where the
	// part is the machine's own: it points at nothing, at this machine, or at an address no other mesh machine has (the
	// resolver a VPN client wrote in). Filled by the node-engine from its tunnel's peers, so the watchman needs no
	// address of any machine (issue 774).
	Targets []string  `json:"targets,omitempty"`
	Since   time.Time `json:"since"`
	Streak  int       `json:"streak,omitempty"`
}

// UnitsHealth is the machine's service managers (issue 315).
type UnitsHealth struct {
	// State is running, degraded or unknown.
	State string `json:"state"`
	// Failed is every unit failed on two looks in a row that no module places.
	Failed []FailedUnit `json:"failed"`
	// Unread names a manager that could not be read, with why; Away an account's own manager not running.
	Unread []string `json:"unread,omitempty"`
	Away   []string `json:"away,omitempty"`
}

// FailedUnit is one failed unit no module places.
type FailedUnit struct {
	Unit string `json:"unit"`
	// Scope is system, or user: an account's own manager.
	Scope    string    `json:"scope"`
	Load     string    `json:"load,omitempty"`
	Result   string    `json:"result,omitempty"`
	Resource string    `json:"resource,omitempty"`
	Since    time.Time `json:"since"`
}

// PeerSeen is one peer of the mesh's tunnel: its public key, its allowed addresses, and whether this machine saw a
// handshake with it within five minutes.
type PeerSeen struct {
	Key       string   `json:"key"`
	Addresses []string `json:"addresses,omitempty"`
	Seen      bool     `json:"seen"`
}

// NodeAccount is one account marked relogin needed, with the module whose resource it is and why.
type NodeAccount struct {
	Account  string    `json:"account"`
	Module   string    `json:"module,omitempty"`
	Resource string    `json:"resource,omitempty"`
	Reason   string    `json:"reason"`
	Since    time.Time `json:"since"`
}

// NodeDirectory is one directory used as found: the module declaring it, its resource and path, and why.
type NodeDirectory struct {
	Module   string    `json:"module"`
	Resource string    `json:"resource"`
	Path     string    `json:"path"`
	Reason   string    `json:"reason"`
	Since    time.Time `json:"since"`
}

// EncodeNodeHealth is h as JSON at its version.
func EncodeNodeHealth(h NodeHealth) ([]byte, error) {
	h.V = NodeHealthVersion
	if h.Node == "" {
		return nil, errors.New("a node's health report names its node")
	}
	return json.Marshal(h)
}

// DecodeNodeHealth reads a report.
func DecodeNodeHealth(raw []byte) (NodeHealth, error) {
	var h NodeHealth
	if err := json.Unmarshal(raw, &h); err != nil {
		return NodeHealth{}, fmt.Errorf("a node's health report is not readable: %w", err)
	}
	if h.V != NodeHealthVersion {
		return NodeHealth{}, fmt.Errorf("a node's health report of version %d, and this reads %d", h.V, NodeHealthVersion)
	}
	return h, nil
}

// SameNodeHealth says two reports say the same, whenever each was written.
func SameNodeHealth(a, b NodeHealth) bool {
	a.At, b.At = time.Time{}, time.Time{}
	ra, errA := json.Marshal(a)
	rb, errB := json.Marshal(b)
	return errA == nil && errB == nil && bytes.Equal(ra, rb)
}