update: add status metrics for update progress

src/configs/bigbox-v1-cm-2.json

@@ -1582,6 +1582,10 @@
         }
       },
       "pipelines": {
+        "component_update_lifecycle": {
+          "protocol": "http",
+          "process": []
+        },
         "wann_type": {
           "template": "any_change"
         },

src/services/metrics.lua

@@ -343,9 +343,9 @@ local function handle_metric(msg)
 	if not metric_name then return end
 
 	local pipe_cfg = State.pipelines_map[metric_name]
+	local payload = msg.payload
 	if not pipe_cfg then return end -- no matching pipeline, drop silently
 
-	local payload = msg.payload
 	if type(payload) ~= 'table' then return end
 
 	local value = payload.value

src/services/update/component_watch.lua

@@ -8,6 +8,7 @@ local fibers       = require 'fibers'
 local scoped_work  = require 'devicecode.support.scoped_work'
 local bus_cleanup  = require 'devicecode.support.bus_cleanup'
 local topics       = require 'services.update.topics'
+local queue        = require 'devicecode.support.queue'
 
 local M = {}
 
@@ -45,9 +46,22 @@ local function open_watches(scope, conn, components, queue_len)
 	return watches, nil
 end
 
+local function report_component_fact(params, component, ev)
+	if not params.events_tx then
+		return true, nil
+	end
+	return queue.try_admit_required(params.events_tx, {
+		kind = 'component_fact_changed',
+		component = component,
+		payload = ev.payload,
+		origin = ev.origin,
+	}, 'update_component_fact_changed_admission_failed')
+end
+
 local function watch_loop(scope, params)
 	local observer = assert(params.observer, 'component_watch observer required')
-	local watches, err = open_watches(scope, assert(params.conn, 'component_watch conn required'), params.components, params.queue_len)
+	local conn = assert(params.conn, 'component_watch conn required')
+	local watches, err = open_watches(scope, conn, params.components, params.queue_len)
 	if not watches then error(err or 'component_watch_open_failed', 0) end
 
 	while true do
@@ -62,7 +76,9 @@ local function watch_loop(scope, params)
 			return { role = 'update_component_watch', reason = recv_err or 'component_watch_closed' }
 		end
 		if ev.op == 'retain' or ev.event == 'retain' or ev.type == 'retain' or ev.kind == 'retain' then
-			observer:update_component(component, ev.payload, ev.origin)
+			local ok_update, update_err = observer:update_component(component, ev.payload, ev.origin)
+			if ok_update == nil then error(update_err or 'component_observer_update_failed', 0) end
+			report_component_fact(params, component, ev)
 		elseif ev.op == 'unretain' or ev.event == 'unretain' or ev.type == 'unretain' or ev.kind == 'unretain' then
 			observer:remove_component(component, 'component_unretained')
 		end
@@ -87,6 +103,7 @@ function M.start(scope, params)
 				observer = params.observer,
 				components = components,
 				queue_len = params.queue_len,
+				events_tx = params.events_tx,
 			})
 		end,
 		report = params.report,

src/services/update/lifecycle_metrics.lua

@@ -0,0 +1,108 @@
+-- services/update/lifecycle_metrics.lua
+--
+-- Small publisher helper for component update lifecycle metrics. The Update
+-- service derives lifecycle records from Device component facts, then calls
+-- emit() with the phase it wants to publish.
+
+local M = {}
+
+M.METRIC_NAME = 'component_update_lifecycle'
+
+local METRIC_TOPIC = { 'obs', 'v1', 'update', 'metric', M.METRIC_NAME }
+
+-- Metric payload shape consumed by services.metrics:
+-- {
+--   value = 'started' | 'staged' | 'completed' | 'failed' | 'cancelled',
+--   namespace = { '<component>', 'lifecycle', '<job_id>', '<phase>' },
+--   job_id = '<original job id>', component = '<component id>', ...
+-- }
+-- The namespace is later used as a metrics topic suffix, so each segment must
+-- avoid separators, whitespace, and dashes.
+
+---Return a string safe to use as one metric namespace segment.
+---This is a topic/namespace normaliser, not a security boundary: it preserves
+---letters, digits, and '_', replaces every other run with '_', trims edge
+---underscores, and falls back to 'unknown' if nothing usable remains.
+---@param value any Original component or job identifier.
+---@return string token Namespace-safe token.
+local function safe_token(value)
+	local s = tostring(value or '')
+	s = s:gsub('[^%w_]', '_')
+	s = s:gsub('_+', '_')
+	s = s:gsub('^_+', ''):gsub('_+$', '')
+	if s == '' then return 'unknown' end
+	return s
+end
+
+---@param record table|nil Lifecycle record built from Device component facts.
+---@return string|nil component Component id when present and non-empty.
+local function component_of(record)
+	local component = type(record) == 'table' and record.component or nil
+	if type(component) ~= 'string' or component == '' then return nil end
+	return component
+end
+
+---Build the metric payload expected by services.metrics.
+---@param record table Lifecycle record with job_id/component/state/error fields.
+---@param phase string Lifecycle phase to publish.
+---@param extra table|nil Additional local-observability fields.
+---@return table|nil payload Metric payload, or nil on invalid input.
+---@return string|nil err Error code when payload is nil.
+local function payload(record, phase, extra)
+	local component = component_of(record)
+	if not component then return nil, 'component_required' end
+	if type(phase) ~= 'string' or phase == '' then return nil, 'phase_required' end
+
+	local out = {
+		value = phase,
+		namespace = { safe_token(component), 'lifecycle', safe_token(record.job_id), safe_token(phase) },
+		job_id = record.job_id,
+		component = component,
+		state = record.state,
+		error = record.error,
+	}
+	for k, v in pairs(extra or {}) do
+		out[k] = v
+	end
+	return out, nil
+end
+
+---@param conn table|nil Bus connection fallback, used when no service_base object exists.
+---@param svc table|nil service_base-like object with obs_metric().
+---@param p table Metric payload from payload().
+---@return boolean|nil ok True when published.
+---@return string|nil err Error code when no sink is available.
+local function publish(conn, svc, p)
+	if svc and type(svc.obs_metric) == 'function' then
+		svc:obs_metric(M.METRIC_NAME, p)
+		return true, nil
+	end
+
+	if conn and type(conn.retain) == 'function' then
+		conn:retain(METRIC_TOPIC, p)
+		return true, nil
+	end
+
+	return nil, 'metric_sink_unavailable'
+end
+
+---@param value any
+---@return string
+function M.safe_token(value)
+	return safe_token(value)
+end
+
+---@param conn table|nil
+---@param svc table|nil
+---@param record table Lifecycle record derived from a Device component fact.
+---@param phase string
+---@param extra table|nil
+---@return boolean|nil ok
+---@return string|nil err
+function M.emit(conn, svc, record, phase, extra)
+	local p, err = payload(record, phase, extra)
+	if not p then return nil, err end
+	return publish(conn, svc, p)
+end
+
+return M