fix: Fixes to running on kubernetes

server/internal/api/handlers.go

@@ -20,6 +20,9 @@ func actorFromCtx(c *gin.Context) string {
 }
 
 func RegisterRoutes(r *gin.Engine) {
+	r.GET("/healthz", handleHealthz)
+	r.GET("/readyz", handleReadyz)
+
 	r.GET("/install", handleInstallScript)
 	r.GET("/install.ps1", handleInstallScriptWindows)
 	r.GET("/update", handleUpdateScript)

server/internal/api/health.go

@@ -0,0 +1,61 @@
+package api
+
+import (
+	"context"
+	"net/http"
+	"time"
+
+	"gitea.hostxtra.co.uk/mrhid6/vantage/server/internal/auth"
+	"gitea.hostxtra.co.uk/mrhid6/vantage/server/internal/db"
+	"github.com/gin-gonic/gin"
+)
+
+// The two probes answer different questions on purpose.
+//
+// /healthz is liveness: the process is up and serving. It touches nothing
+// external, because a Mongo outage must not make Kubernetes restart every
+// server pod — a restart loop cannot fix someone else's database, and it
+// destroys every open command stream and console session on the way.
+//
+// /readyz is readiness: this pod can serve a request end to end, which needs
+// both Mongo and Redis. A failing readiness probe pulls the pod out of the
+// Service and leaves it running, which is the behaviour that matters during a
+// dependency blip.
+//
+// Both sit outside /api, so neither the session middleware nor the licence
+// gate applies. Neither reveals anything beyond up or down.
+
+const probeTimeout = 2 * time.Second
+
+func handleHealthz(c *gin.Context) {
+	c.JSON(http.StatusOK, gin.H{"status": "ok"})
+}
+
+func handleReadyz(c *gin.Context) {
+	ctx, cancel := context.WithTimeout(c.Request.Context(), probeTimeout)
+	defer cancel()
+
+	checks := gin.H{"mongo": "ok", "redis": "ok"}
+	ready := true
+
+	if db.Client == nil {
+		checks["mongo"] = "not initialised"
+		ready = false
+	} else if err := db.Client.Ping(ctx, nil); err != nil {
+		checks["mongo"] = "unreachable"
+		ready = false
+	}
+
+	if err := auth.PingRedis(ctx); err != nil {
+		checks["redis"] = "unreachable"
+		ready = false
+	}
+
+	status := http.StatusOK
+	state := "ok"
+	if !ready {
+		status = http.StatusServiceUnavailable
+		state = "unready"
+	}
+	c.JSON(status, gin.H{"status": state, "checks": checks})
+}

server/internal/api/workflows.go

@@ -5,7 +5,6 @@ import (
 	"fmt"
 	"io"
 	"net/http"
-	"os"
 	"regexp"
 	"strconv"
 	"strings"
@@ -50,23 +49,42 @@ func getServerRunLog(c *gin.Context) {
 		c.JSON(http.StatusBadRequest, gin.H{"error": "invalid id"})
 		return
 	}
-	path := services.ServerRunLogPath(runID, serverID)
-	b, err := os.ReadFile(path)
-	if err != nil {
+	if !services.HasServerRunLog(runID, serverID) {
 		c.JSON(http.StatusNotFound, gin.H{"error": "no logs"})
 		return
 	}
-	c.Data(http.StatusOK, "text/plain; charset=utf-8", b)
+
+	c.Header("Content-Type", "text/plain; charset=utf-8")
+	c.Status(http.StatusOK)
+
+	// Streamed in pages rather than read whole. A log capped at 200k lines is
+	// tens of megabytes, and holding that in memory per concurrent download is
+	// how one curious user takes a pod down.
+	var after int64
+	for {
+		lines, last, err := services.ReadServerRunLog(runID, serverID, after, logPageSize)
+		if err != nil || len(lines) == 0 {
+			return
+		}
+		for _, l := range lines {
+			_, _ = c.Writer.WriteString(l)
+			_, _ = c.Writer.WriteString("\n")
+		}
+		c.Writer.Flush()
+		after = last
+	}
 }
 
+// logPageSize bounds one read of the log store. Large enough that a normal log
+// is one or two queries, small enough that no single response buffers much.
+const logPageSize = 2000
+
 func streamServerRunLog(c *gin.Context) {
 	runID, serverID := c.Param("runId"), c.Param("serverId")
 	if !uuidLike.MatchString(runID) || !uuidLike.MatchString(serverID) {
 		c.JSON(http.StatusBadRequest, gin.H{"error": "invalid id"})
 		return
 	}
-	path := services.ServerRunLogPath(runID, serverID)
-
 	c.Writer.Header().Set("Content-Type", "text/event-stream")
 	c.Writer.Header().Set("Cache-Control", "no-cache")
 	c.Writer.Header().Set("Connection", "keep-alive")
@@ -78,29 +96,27 @@ func streamServerRunLog(c *gin.Context) {
 		return
 	}
 
-	var offset int64
+	// The cursor is a sequence number now, not a byte offset: lines come from
+	// the log store, which any pod can read, rather than from a file only this
+	// one has.
+	var after int64
 	sendNew := func() {
-		f, err := os.Open(path)
-		if err != nil {
-			return
-		}
-		defer f.Close()
-		if _, err := f.Seek(offset, 0); err != nil {
-			return
-		}
-		buf := make([]byte, 32*1024)
 		for {
-			n, _ := f.Read(buf)
-			if n <= 0 {
-				break
+			lines, last, err := services.ReadServerRunLog(runID, serverID, after, logPageSize)
+			if err != nil || len(lines) == 0 {
+				return
 			}
-			offset += int64(n)
-
-			for _, line := range splitSSE(buf[:n]) {
-				_, _ = c.Writer.WriteString("data: " + line + "\n")
+			after = last
+			for _, line := range lines {
+				for _, part := range splitSSE([]byte(line)) {
+					_, _ = c.Writer.WriteString("data: " + part + "\n")
+				}
+				_, _ = c.Writer.WriteString("\n")
 			}
-			_, _ = c.Writer.WriteString("\n")
 			flusher.Flush()
+			if len(lines) < logPageSize {
+				return
+			}
 		}
 	}