Respond to Steven's feedback

ammario · ammario · commit 96f27c6ea7b3 · 2022-08-30T23:38:19.000Z
diff --git a/coderd/database/migrations/000039_agent_stats.up.sql b/coderd/database/migrations/000039_agent_stats.up.sql
@@ -1,5 +1,5 @@
 CREATE TABLE agent_stats (
-	id text NOT NULL,
+	id uuid NOT NULL,
 	PRIMARY KEY (id),
     created_at timestamptz NOT NULL,
     user_id uuid NOT NULL,
diff --git a/coderd/metrics.go b/coderd/metrics.go
@@ -76,7 +76,7 @@ func (api *API) daus(rw http.ResponseWriter, r *http.Request) {
 	httpapi.Write(rw, http.StatusOK, resp)
 }
 
-const AgentStatIntervalEnv = "CODER_AGENT_STAT_INTERVAL"
+const AgentStatIntervalEnv = "CODER_AGENT_STAT_INTERVAL_MS"
 
 func (api *API) workspaceAgentReportStats(rw http.ResponseWriter, r *http.Request) {
 	api.websocketWaitMutex.Lock()
@@ -177,6 +177,7 @@ func (api *API) workspaceAgentReportStats(rw http.ResponseWriter, r *http.Reques
 		var insert = !reflect.DeepEqual(lastReport, rep)
 
 		api.Logger.Debug(ctx, "read stats report",
+			slog.F("interval", interval),
 			slog.F("agent", workspaceAgent.ID),
 			slog.F("resource", resource.ID),
 			slog.F("workspace", workspace.ID),
diff --git a/codersdk/metrics.go b/codersdk/metrics.go
@@ -55,7 +55,11 @@ func (c *Client) AgentReportStats(
 	go func() {
 		defer close(doneCh)
 
-		for r := retry.New(time.Second, time.Hour); r.Wait(ctx); {
+		// If the agent connection succeeds for a while, then fails, then succeeds
+		// for a while (etc.) the retry may hit the maximum. This is a normal
+		// case for long-running agents that experience coderd upgrades, so
+		// we use a short maximum retry limit.
+		for r := retry.New(time.Second, time.Minute); r.Wait(ctx); {
 			err = func() error {
 				conn, res, err := websocket.Dial(ctx, serverURL.String(), &websocket.DialOptions{
 					HTTPClient: httpClient,
