Merge branch '21283-callable-api-module'

[arvados.git] / services / ws / router.go

diff --git a/services/ws/router.go b/services/ws/router.go
index 69654cbe529affde9ff46b7906ff6392240096ef..878c282f8a6c57f17192b777faba760485757b86 100644 (file)
--- a/services/ws/router.go
+++ b/services/ws/router.go
@@ -1,74 +1,112 @@
-package main
+// Copyright (C) The Arvados Authors. All rights reserved.
+//
+// SPDX-License-Identifier: AGPL-3.0
+
+package ws
 
 import (
-       "encoding/json"
-       "log"
+       "io"
        "net/http"
        "sync"
+       "sync/atomic"
        "time"
 
-       "git.curoverse.com/arvados.git/sdk/go/arvados"
+       "git.arvados.org/arvados.git/sdk/go/arvados"
+       "git.arvados.org/arvados.git/sdk/go/ctxlog"
+       "git.arvados.org/arvados.git/sdk/go/health"
+       "github.com/prometheus/client_golang/prometheus"
+       "github.com/sirupsen/logrus"
        "golang.org/x/net/websocket"
 )
 
+type wsConn interface {
+       io.ReadWriter
+       Request() *http.Request
+       SetReadDeadline(time.Time) error
+       SetWriteDeadline(time.Time) error
+}
+
 type router struct {
-       Config *Config
+       client         *arvados.Client
+       cluster        *arvados.Cluster
+       eventSource    eventSource
+       newPermChecker func() permChecker
 
-       eventSource eventSource
-       mux         *http.ServeMux
-       setupOnce   sync.Once
+       handler   *handler
+       mux       *http.ServeMux
+       setupOnce sync.Once
+       done      chan struct{}
+       reg       *prometheus.Registry
 }
 
 func (rtr *router) setup() {
+       mSockets := prometheus.NewGaugeVec(prometheus.GaugeOpts{
+               Namespace: "arvados",
+               Subsystem: "ws",
+               Name:      "sockets",
+               Help:      "Number of connected sockets",
+       }, []string{"version"})
+       rtr.reg.MustRegister(mSockets)
+
+       rtr.handler = &handler{
+               PingTimeout: time.Duration(rtr.cluster.API.SendTimeout),
+               QueueSize:   rtr.cluster.API.WebsocketClientEventQueue,
+       }
        rtr.mux = http.NewServeMux()
-       rtr.mux.Handle("/websocket", rtr.makeServer(NewSessionV0))
-       rtr.mux.Handle("/arvados/v1/events.ws", rtr.makeServer(NewSessionV1))
+       rtr.mux.Handle("/websocket", rtr.makeServer(newSessionV0, mSockets.WithLabelValues("0")))
+       rtr.mux.Handle("/arvados/v1/events.ws", rtr.makeServer(newSessionV1, mSockets.WithLabelValues("1")))
+       rtr.mux.Handle("/_health/", &health.Handler{
+               Token:  rtr.cluster.ManagementToken,
+               Prefix: "/_health/",
+               Routes: health.Routes{
+                       "db": rtr.eventSource.DBHealth,
+               },
+               Log: func(r *http.Request, err error) {
+                       if err != nil {
+                               ctxlog.FromContext(r.Context()).WithError(err).Error("error")
+                       }
+               },
+       })
 }
 
-func (rtr *router) makeServer(newSession func(wsConn, arvados.Client) (session, error)) *websocket.Server {
-       handler := &handler{
-               Client:      rtr.Config.Client,
-               PingTimeout: rtr.Config.PingTimeout.Duration(),
-               QueueSize:   rtr.Config.ClientEventQueue,
-               NewSession:  newSession,
-       }
+func (rtr *router) makeServer(newSession sessionFactory, gauge prometheus.Gauge) *websocket.Server {
+       var connected int64
        return &websocket.Server{
                Handshake: func(c *websocket.Config, r *http.Request) error {
                        return nil
                },
                Handler: websocket.Handler(func(ws *websocket.Conn) {
-                       sink := rtr.eventSource.NewSink()
-                       handler.Handle(ws, sink.Channel())
-                       sink.Stop()
+                       t0 := time.Now()
+                       logger := ctxlog.FromContext(ws.Request().Context())
+                       atomic.AddInt64(&connected, 1)
+                       gauge.Set(float64(atomic.LoadInt64(&connected)))
+
+                       stats := rtr.handler.Handle(ws, logger, rtr.eventSource,
+                               func(ws wsConn, sendq chan<- interface{}) (session, error) {
+                                       return newSession(ws, sendq, rtr.eventSource.DB(), rtr.newPermChecker(), rtr.client)
+                               })
+
+                       logger.WithFields(logrus.Fields{
+                               "elapsed": time.Now().Sub(t0).Seconds(),
+                               "stats":   stats,
+                       }).Info("client disconnected")
                        ws.Close()
+                       atomic.AddInt64(&connected, -1)
+                       gauge.Set(float64(atomic.LoadInt64(&connected)))
                }),
        }
 }
 
 func (rtr *router) ServeHTTP(resp http.ResponseWriter, req *http.Request) {
        rtr.setupOnce.Do(rtr.setup)
-       t0 := time.Now()
-       reqLog(map[string]interface{}{
-               "Connect":         req.RemoteAddr,
-               "RemoteAddr":      req.RemoteAddr,
-               "X-Forwarded-For": req.Header.Get("X-Forwarded-For"),
-               "Time":            t0.UTC(),
-       })
        rtr.mux.ServeHTTP(resp, req)
-       t1 := time.Now()
-       reqLog(map[string]interface{}{
-               "Disconnect":      req.RemoteAddr,
-               "RemoteAddr":      req.RemoteAddr,
-               "X-Forwarded-For": req.Header.Get("X-Forwarded-For"),
-               "Time":            t1.UTC(),
-               "Elapsed":         time.Now().Sub(t0).Seconds(),
-       })
 }
 
-func reqLog(m map[string]interface{}) {
-       j, err := json.Marshal(m)
-       if err != nil {
-               log.Fatal(err)
-       }
-       log.Print(string(j))
+func (rtr *router) CheckHealth() error {
+       rtr.setupOnce.Do(rtr.setup)
+       return rtr.eventSource.DBHealth()
+}
+
+func (rtr *router) Done() <-chan struct{} {
+       return rtr.done
 }