From 5e338b0dae7e7e79978726e5fdd39e4eb6badcba Mon Sep 17 00:00:00 2001
From: David Colburn <xero73@gmail.com>
Date: Fri, 31 May 2024 20:22:50 -0700
Subject: [PATCH] clean up zombies (#686)

* attempts

* reap zombies with tini

* logs

* clean up

---------

Co-authored-by: David Zhao <dz@livekit.io>
---
 build/egress/Dockerfile    | 20 ++++++++++++++------
 build/egress/entrypoint.sh |  2 +-
 pkg/pipeline/source/web.go | 21 ++++++++++++---------
 pkg/server/server_rpc.go   |  2 ++
 4 files changed, 29 insertions(+), 16 deletions(-)

diff --git a/build/egress/Dockerfile b/build/egress/Dockerfile
index fbfa47b0..2f3c210e 100644
--- a/build/egress/Dockerfile
+++ b/build/egress/Dockerfile
@@ -15,17 +15,18 @@
 FROM livekit/gstreamer:1.22.8-dev
 
 ARG TARGETPLATFORM
+ARG TARGETARCH
+ENV TARGETARCH=${TARGETARCH}
+ENV TARGETPLATFORM=${TARGETPLATFORM}
 
 WORKDIR /workspace
 
 # install go
-RUN if [ "$TARGETPLATFORM" = "linux/arm64" ]; then GOARCH=arm64; else GOARCH=amd64; fi && \
-    wget https://go.dev/dl/go1.22.1.linux-${GOARCH}.tar.gz && \
+RUN wget https://go.dev/dl/go1.22.1.linux-${TARGETARCH}.tar.gz && \
     rm -rf /usr/local/go && \
-    tar -C /usr/local -xzf go1.22.1.linux-${GOARCH}.tar.gz
+    tar -C /usr/local -xzf go1.22.1.linux-${TARGETARCH}.tar.gz
 ENV PATH="/usr/local/go/bin:${PATH}"
 
-
 # download go modules
 COPY go.mod .
 COPY go.sum .
@@ -42,8 +43,13 @@ COPY --from=livekit/egress-templates workspace/build/ cmd/server/templates/
 RUN find cmd/server/templates/ -name *.map | xargs rm
 
 # build
-RUN if [ "$TARGETPLATFORM" = "linux/arm64" ]; then GOARCH=arm64; else GOARCH=amd64; fi && \
-    CGO_ENABLED=1 GOOS=linux GOARCH=${GOARCH} GO111MODULE=on GODEBUG=disablethp=1 go build -a -o egress ./cmd/server
+RUN CGO_ENABLED=1 GOOS=linux GOARCH=${TARGETARCH} GO111MODULE=on GODEBUG=disablethp=1 go build -a -o egress ./cmd/server
+
+# install tini
+ENV TINI_VERSION v0.19.0
+
+ADD https://github.com/krallin/tini/releases/download/${TINI_VERSION}/tini-${TARGETARCH} /tini
+RUN chmod +x /tini
 
 FROM livekit/gstreamer:1.22.8-prod
 
@@ -73,8 +79,10 @@ RUN useradd -ms /bin/bash -g root -G sudo,pulse,pulse-access egress
 RUN mkdir -p home/egress/tmp home/egress/.cache/xdgr && \
     chown -R egress /home/egress
 
+
 # copy files
 COPY --from=0 /workspace/egress /bin/
+COPY --from=0 /tini /tini
 COPY build/egress/entrypoint.sh /
 
 # run
diff --git a/build/egress/entrypoint.sh b/build/egress/entrypoint.sh
index b406a3fc..9e40cd50 100755
--- a/build/egress/entrypoint.sh
+++ b/build/egress/entrypoint.sh
@@ -23,4 +23,4 @@ rm -rf /var/run/pulse /var/lib/pulse /home/egress/.config/pulse /home/egress/.ca
 pulseaudio -D --verbose --exit-idle-time=-1 --disallow-exit
 
 # Run egress service
-exec egress
+exec /tini -- egress
diff --git a/pkg/pipeline/source/web.go b/pkg/pipeline/source/web.go
index bbba4905..c7106871 100644
--- a/pkg/pipeline/source/web.go
+++ b/pkg/pipeline/source/web.go
@@ -43,9 +43,9 @@ const (
 )
 
 type WebSource struct {
-	pulseSink    string
-	xvfb         *exec.Cmd
-	chromeCancel context.CancelFunc
+	pulseSink   string
+	xvfb        *exec.Cmd
+	closeChrome context.CancelFunc
 
 	startRecording chan struct{}
 	endRecording   chan struct{}
@@ -109,10 +109,10 @@ func (s *WebSource) GetEndedAt() int64 {
 }
 
 func (s *WebSource) Close() {
-	if s.chromeCancel != nil {
+	if s.closeChrome != nil {
 		logger.Debugw("closing chrome")
-		s.chromeCancel()
-		s.chromeCancel = nil
+		s.closeChrome()
+		s.closeChrome = nil
 	}
 
 	if s.xvfb != nil {
@@ -254,9 +254,12 @@ func (s *WebSource) launchChrome(ctx context.Context, p *config.PipelineConfig,
 		)
 	}
 
-	allocCtx, _ := chromedp.NewExecAllocator(context.Background(), opts...)
-	chromeCtx, cancel := chromedp.NewContext(allocCtx)
-	s.chromeCancel = cancel
+	allocCtx, allocCancel := chromedp.NewExecAllocator(context.Background(), opts...)
+	chromeCtx, chromeCancel := chromedp.NewContext(allocCtx)
+	s.closeChrome = func() {
+		chromeCancel()
+		allocCancel()
+	}
 
 	chromedp.ListenTarget(chromeCtx, func(ev interface{}) {
 		switch ev := ev.(type) {
diff --git a/pkg/server/server_rpc.go b/pkg/server/server_rpc.go
index 27fa39a2..56b09c8f 100644
--- a/pkg/server/server_rpc.go
+++ b/pkg/server/server_rpc.go
@@ -20,6 +20,7 @@ import (
 	"os"
 	"os/exec"
 	"path"
+	"syscall"
 	"time"
 
 	"google.golang.org/protobuf/encoding/protojson"
@@ -119,6 +120,7 @@ func (s *Server) launchProcess(req *rpc.StartEgressRequest, info *livekit.Egress
 	cmd.Dir = "/"
 	cmd.Stdout = os.Stdout
 	cmd.Stderr = os.Stderr
+	cmd.SysProcAttr = &syscall.SysProcAttr{Setsid: true}
 
 	s.monitor.EgressStarted(req)