history: fix the more nuanced cache-control max-age logic

Dockerfile

@@ -1,4 +1,4 @@
-FROM alpine:3.20.3
+FROM alpine:3.21.0
 
 RUN apk --no-cache upgrade
 RUN apk --no-cache add ca-certificates tzdata zsh jq tmux curl

go.mod

@@ -16,7 +16,7 @@ require (
 	github.com/spf13/cobra v1.8.1
 	github.com/stretchr/testify v1.10.0
 	go.ntppool.org/api v0.3.4
-	go.ntppool.org/common v0.3.0
+	go.ntppool.org/common v0.3.1
 	go.opentelemetry.io/contrib/instrumentation/github.com/labstack/echo/otelecho v0.58.0
 	go.opentelemetry.io/contrib/instrumentation/net/http/httptrace/otelhttptrace v0.58.0
 	go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.58.0
@@ -92,8 +92,8 @@ require (
 	golang.org/x/sys v0.28.0 // indirect
 	golang.org/x/text v0.21.0 // indirect
 	golang.org/x/time v0.8.0 // indirect
-	google.golang.org/genproto/googleapis/api v0.0.0-20241223144023-3abc09e42ca8 // indirect
+	google.golang.org/genproto/googleapis/api v0.0.0-20241230172942-26aa7a208def // indirect
-	google.golang.org/genproto/googleapis/rpc v0.0.0-20241223144023-3abc09e42ca8 // indirect
+	google.golang.org/genproto/googleapis/rpc v0.0.0-20241230172942-26aa7a208def // indirect
 	google.golang.org/grpc v1.69.2 // indirect
 	google.golang.org/protobuf v1.36.1 // indirect
 )

go.sum

@@ -138,8 +138,8 @@ github.com/yuin/goldmark v1.2.1/go.mod h1:3hX8gzYuyVAZsxl0MRgGTJEmQBFcNTphYh9dec
 go.mongodb.org/mongo-driver v1.11.4/go.mod h1:PTSz5yu21bkT/wXpkS7WR5f0ddqw5quethTUn9WM+2g=
 go.ntppool.org/api v0.3.4 h1:KeRyFhIRkjJwZif7hkpqEDEBmukyYGiOi2Fd6j3UzQ0=
 go.ntppool.org/api v0.3.4/go.mod h1:LFLAwnrc/JyjzKnjgf8tCOJhps6oFIjuledS3PCx7xc=
-go.ntppool.org/common v0.3.0 h1:IuSmyjEhI1F3tr5kc5MqlR4cy5y0o5f3EKvC7Koc6rs=
+go.ntppool.org/common v0.3.1 h1:JaJpS3m8oAc9jH0yhHYJnjOC+RUzxx/F+EDe0QON4eQ=
-go.ntppool.org/common v0.3.0/go.mod h1:25pUt3YUusF1MY0nsljjskcMMeTvKZszVvNsubvWhSM=
+go.ntppool.org/common v0.3.1/go.mod h1:1SKjFBH9AL7Fj2S0zy41isHzV6dTC+6yIKD5QDtX8aY=
 go.opentelemetry.io/auto/sdk v1.1.0 h1:cH53jehLUN6UFLY71z+NDOiNJqDdPRaXzTel0sJySYA=
 go.opentelemetry.io/auto/sdk v1.1.0/go.mod h1:3wSPjt5PWp2RhlCcmmOial7AvC4DQqZb7a7wCow3W8A=
 go.opentelemetry.io/contrib/bridges/otelslog v0.8.0 h1:G3sKsNueSdxuACINFxKrQeimAIst0A5ytA2YJH+3e1c=
@@ -246,10 +246,10 @@ golang.org/x/xerrors v0.0.0-20190717185122-a985d3407aa7/go.mod h1:I/5z698sn9Ka8T
 golang.org/x/xerrors v0.0.0-20191011141410-1b5146add898/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
 golang.org/x/xerrors v0.0.0-20191204190536-9bdfabe68543/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
 golang.org/x/xerrors v0.0.0-20200804184101-5ec99f83aff1/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
-google.golang.org/genproto/googleapis/api v0.0.0-20241223144023-3abc09e42ca8 h1:st3LcW/BPi75W4q1jJTEor/QWwbNlPlDG0JTn6XhZu0=
+google.golang.org/genproto/googleapis/api v0.0.0-20241230172942-26aa7a208def h1:0Km0hi+g2KXbXL0+riZzSCKz23f4MmwicuEb00JeonI=
-google.golang.org/genproto/googleapis/api v0.0.0-20241223144023-3abc09e42ca8/go.mod h1:klhJGKFyG8Tn50enBn7gizg4nXGXJ+jqEREdCWaPcV4=
+google.golang.org/genproto/googleapis/api v0.0.0-20241230172942-26aa7a208def/go.mod h1:u2DoMSpCXjrzqLdobRccQMc9wrnMAJ1DLng0a2yqM2Q=
-google.golang.org/genproto/googleapis/rpc v0.0.0-20241223144023-3abc09e42ca8 h1:TqExAhdPaB60Ux47Cn0oLV07rGnxZzIsaRhQaqS666A=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20241230172942-26aa7a208def h1:4P81qv5JXI/sDNae2ClVx88cgDDA6DPilADkG9tYKz8=
-google.golang.org/genproto/googleapis/rpc v0.0.0-20241223144023-3abc09e42ca8/go.mod h1:lcTa1sDdWEIHMWlITnIczmw5w60CF9ffkb8Z+DVmmjA=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20241230172942-26aa7a208def/go.mod h1:bdAgzvd4kFrpykc5/AC2eLUiegK9T/qxZHD4hXYf/ho=
 google.golang.org/grpc v1.69.2 h1:U3S9QEtbXC0bYNvRtcoklF3xGtLViumSYxWykJS+7AU=
 google.golang.org/grpc v1.69.2/go.mod h1:vyjdE6jLBI76dgpDojsFGNaHlxdjXN9ghpnd2o7JGZ4=
 google.golang.org/protobuf v1.26.0-rc.1/go.mod h1:jlhhOSvTdKEhbULTjvd4ARK9grFBp09yW+WbY/TyQbw=

server/history.go

@@ -73,7 +73,7 @@ func (srv *Server) getHistoryParameters(ctx context.Context, c echo.Context) (hi
 
 	monitorParam := c.QueryParam("monitor")
 
-	var monitorID uint32 = 0
+	var monitorID uint32
 	switch monitorParam {
 	case "":
 		name := "recentmedian.scores.ntp.dev"
@@ -145,7 +145,8 @@ func (srv *Server) history(c echo.Context) error {
 	ctx, span := tracing.Tracer().Start(c.Request().Context(), "history")
 	defer span.End()
 
-	// just cache for a short time by default
+	// set a reasonable default cache time; adjusted later for
+	// happy path common responses
 	c.Response().Header().Set("Cache-Control", "public,max-age=240")
 
 	mode := paramHistoryMode(c.Param("mode"))
@@ -223,7 +224,6 @@ func (srv *Server) history(c echo.Context) error {
 	default:
 		return c.String(http.StatusNotFound, "not implemented")
 	}
-
 }
 
 func (srv *Server) historyJSON(ctx context.Context, c echo.Context, server ntpdb.Server, history *logscores.LogScoreHistory) error {
@@ -317,21 +317,16 @@ func (srv *Server) historyJSON(ctx context.Context, c echo.Context, server ntpdb
 		}
 	}
 
-	if len(history.LogScores) == 0 ||
+	setHistoryCacheControl(c, history)
-		history.LogScores[len(history.LogScores)-1].Ts.After(time.Now().Add(-8*time.Hour)) {
-		// cache for longer if data hasn't updated for a while
-		c.Request().Header.Set("Cache-Control", "s-maxage=3600,max-age=1800")
-	} else {
-		c.Request().Header.Set("Cache-Control", "s-maxage=300,max-age=240")
-	}
 
 	return c.JSON(http.StatusOK, res)
-
 }
 
 func (srv *Server) historyCSV(ctx context.Context, c echo.Context, history *logscores.LogScoreHistory) error {
 	log := logger.Setup()
 	ctx, span := tracing.Tracer().Start(ctx, "history.csv")
+	defer span.End()
+
 	b := bytes.NewBuffer([]byte{})
 	w := csv.NewWriter(b)
 
@@ -342,7 +337,11 @@ func (srv *Server) historyCSV(ctx context.Context, c echo.Context, history *logs
 		return s
 	}
 
-	w.Write([]string{"ts_epoch", "ts", "offset", "step", "score", "monitor_id", "monitor_name", "leap", "error"})
+	err := w.Write([]string{"ts_epoch", "ts", "offset", "step", "score", "monitor_id", "monitor_name", "leap", "error"})
+	if err != nil {
+		log.ErrorContext(ctx, "could not write csv header", "err", err)
+		return err
+	}
 	for _, l := range history.LogScores {
 		// log.Debug("csv line", "id", l.ID, "n", i)
 
@@ -381,16 +380,31 @@ func (srv *Server) historyCSV(ctx context.Context, c echo.Context, history *logs
 	w.Flush()
 	if err := w.Error(); err != nil {
 		log.ErrorContext(ctx, "could not flush csv", "err", err)
-		span.End()
 		return c.String(http.StatusInternalServerError, "csv error")
 	}
 
 	// log.Info("entries", "count", len(history.LogScores), "out_bytes", b.Len())
 
-	c.Response().Header().Set("Cache-Control", "s-maxage=150,max-age=120")
+	setHistoryCacheControl(c, history)
+
 	c.Response().Header().Set("Content-Disposition", "inline")
 	// Chrome and Firefox force-download text/csv files, so use text/plain
 	// https://bugs.chromium.org/p/chromium/issues/detail?id=152911
 	return c.Blob(http.StatusOK, "text/plain", b.Bytes())
-
+}
+
+func setHistoryCacheControl(c echo.Context, history *logscores.LogScoreHistory) {
+	hdr := c.Response().Header()
+	if len(history.LogScores) == 0 ||
+		// cache for longer if data hasn't updated for a while; or we didn't
+		// find any.
+		(time.Now().Add(-8 * time.Hour).After(history.LogScores[len(history.LogScores)-1].Ts)) {
+		hdr.Set("Cache-Control", "s-maxage=3600,max-age=1800")
+	} else {
+		if len(history.LogScores) == 1 {
+			hdr.Set("Cache-Control", "s-maxage=60,max-age=35")
+		} else {
+			hdr.Set("Cache-Control", "s-maxage=240,max-age=120")
+		}
+	}
 }

server/server.go

@@ -76,7 +76,7 @@ func NewServer(ctx context.Context, configFile string) (*Server, error) {
 		Environment: conf.DeploymentMode(),
 	})
 	if err != nil {
-		return nil, err
+		return nil, fmt.Errorf("tracing init: %w", err)
 	}
 
 	srv.tpShutdown = append(srv.tpShutdown, tpShutdown)
@@ -274,22 +274,22 @@ func (srv *Server) userCountryData(c echo.Context) error {
 		UserCountry: data,
 		ZoneStats:   zoneStats,
 	})
-
 }
 
 func healthHandler(srv *Server, log *slog.Logger) func(w http.ResponseWriter, req *http.Request) {
-
 	return func(w http.ResponseWriter, req *http.Request) {
-
 		ctx := req.Context()
 		ctx, cancel := context.WithTimeout(ctx, 5*time.Second)
 		defer cancel()
 		g, ctx := errgroup.WithContext(ctx)
 
 		stats := srv.db.Stats()
+		if stats.OpenConnections > 5 {
 			log.InfoContext(ctx, "health requests", "url", req.URL.String(), "stats", stats)
+		}
 
-		reset, err := strconv.ParseBool(req.URL.Query().Get("reset"))
+		if resetParam := req.URL.Query().Get("reset"); resetParam != "" {
+			reset, err := strconv.ParseBool(resetParam)
 			log.InfoContext(ctx, "db reset request", "err", err, "reset", reset)
 
 			if err == nil && reset {
@@ -297,6 +297,7 @@ func healthHandler(srv *Server, log *slog.Logger) func(w http.ResponseWriter, re
 				srv.db.SetMaxIdleConns(0)
 				srv.db.SetConnMaxLifetime(5 * time.Second)
 			}
+		}
 
 		g.Go(func() error {
 			err := srv.ch.Scores.Ping(ctx)
@@ -316,13 +317,19 @@ func healthHandler(srv *Server, log *slog.Logger) func(w http.ResponseWriter, re
 			return nil
 		})
 
-		err = g.Wait()
+		err := g.Wait()
 		if err != nil {
 			w.WriteHeader(http.StatusServiceUnavailable)
-			w.Write([]byte("db ping err"))
+			_, err = w.Write([]byte("db ping err"))
+			if err != nil {
+				log.ErrorContext(ctx, "could not write response", "err", err)
+			}
 			return
 		}
 		w.WriteHeader(http.StatusOK)
-		w.Write([]byte("ok"))
+		_, err = w.Write([]byte("ok"))
+		if err != nil {
+			log.ErrorContext(ctx, "could not write response", "err", err)
+		}
 	}
 }