Merge pull request #88 from tbflw/feature/run_non_root_and_log_startup

Feature/run non root and log startup

.dockerignore

@@ -0,0 +1,2 @@
+scripts
+test

.gitignore

@@ -0,0 +1 @@
+.idea

Makefile

@@ -2,6 +2,7 @@ GOLANG_VERSION=1.12
 VERSION=snapshot
 DOCKER_USER=
 DOCKER_PASSWORD=
+DOCKER_REPO=thecodingmachine
 
 # generate documentation.
 doc:
@@ -27,7 +28,7 @@ tests:
 # build Docker image.
 image:
 	docker build -t thecodingmachine/gotenberg:base -f build/base/Dockerfile .
-	docker build --build-arg GOLANG_VERSION=$(GOLANG_VERSION) --build-arg VERSION=$(VERSION) -t thecodingmachine/gotenberg:$(VERSION) -f build/package/Dockerfile .
+	docker build --build-arg GOLANG_VERSION=$(GOLANG_VERSION) --build-arg VERSION=$(VERSION) -t $(DOCKER_REPO)/gotenberg:$(VERSION) -f build/package/Dockerfile .
 
 # start the API using previously built Docker image.
 gotenberg:

build/base/Dockerfile

@@ -1,4 +1,4 @@
-FROM debian:9.5-slim
+FROM debian:9-slim
 
 # |--------------------------------------------------------------------------
 # | Common libraries
@@ -91,3 +91,16 @@ RUN apt-get install -y \
     fonts-unfonts-core
 
 COPY build/base/fonts.conf /etc/fonts/conf.d/100-gotenberg.conf
+
+# |--------------------------------------------------------------------------
+# | Default user
+# |--------------------------------------------------------------------------
+# |
+# | TODO find a correct description for why we're doing this.
+# |
+
+RUN groupadd --gid 1001 gotenberg \
+  && useradd --uid 1001 --gid gotenberg --shell /bin/bash --no-create-home gotenberg \
+  && mkdir /gotenberg \
+  && chown gotenberg: /gotenberg
+

build/docs/content/01-install.md

@@ -10,6 +10,8 @@ You may start it with:
 $ docker run --rm -p 3000:3000 thecodingmachine/gotenberg:5
 ```
 
+All processes in the docker container runs as a dedicated non-root user called `gotenberg` with user id `1001` from the working dir `/gotenberg`.
+
 > The API will be available at [http://localhost:3000](http://localhost:3000).
 
 ## Docker Compose
@@ -38,5 +40,12 @@ Otherwise the API will not be able to launch Google Chrome and LibreOffice (unoc
 
 > The more resources are granted, the quicker will be the conversions.
 
+Also, in the deployment spec of the pod, specify the uid `1001` of the user `gotenberg`: 
+```
+        securityContext:
+          privileged: false
+          runAsUser: 1001
+```
+
 In the following examples, we will assume your
 Gotenberg API is available at [http://localhost:3000](http://localhost:3000).

build/docs/content/03-environment-variables.md

@@ -8,7 +8,7 @@ You may customize the API behaviour thanks to environment variables.
 
 In order to save some resources, the Gotenberg image accepts the environment variable `DISABLE_GOOGLE_CHROME`.
 
-It takes the strings `"0"` or `"1"` as value.
+It takes the strings `"0"` or `"1"` as value where `1` means `true`
 
 > If Google Chrome is disabled, the following conversions will **not** be available anymore:
 > [HTML](#html), [URL](#url) and [Markdown](#markdown)
@@ -37,12 +37,19 @@ By default, the API will add a log entry when the [healthcheck endpoint](#ping)
 
 You may turn off this logging so as to avoid unnecessary entries in your logs with the environment variable `DISABLE_HEALTHCHECK_LOGGING`.
 
-This environment variable operates in the same manner as the `DISABLE_GOOGLE_CHROME` and `DISABLE_UNOCONV` variables operate in that it accepts the strings `"0"` or `"1"` as values. 
+This environment variable operates in the same manner as the `DISABLE_GOOGLE_CHROME` and `DISABLE_UNOCONV` variables operate in that it accepts the strings `"0"` or `"1"` as values, where `1` is enabled. 
-
 ## Default listen port
 
 By default, the API will listen on port `3000`. For most use cases this is perfectly fine, but at times there may be cases where you need to change this due to port conflicts.
 
 You may customize this port location with the environment variable `DEFAULT_LISTEN_PORT`.
 
-This environment variable accepts any string that can be turned into a port number (e.g., the string `"0"` up to the string `"65535"`).
+This environment variable accepts any string that can be turned into a port number (e.g., the string `"0"` up to the string `"65535"`).
+
+## Debug logging of process startup
+
+By default, stdout and stderr messages from the started processes are disabled.
+
+You may enable some debug logging from starting the process by setting the environment variable `DEBUG_PROCESS_STARTUP`.
+
+This environment variable operates in the same manner as the `DISABLE_GOOGLE_CHROME` and `DISABLE_UNOCONV` variables operate in that it accepts the strings `"0"` or `"1"` as values, where `1` means `true`. 

build/package/Dockerfile

@@ -19,7 +19,11 @@ ENV GOOS=linux \
 WORKDIR /gotenberg
 
 # Copy our source code.
-COPY . .
+COPY internal ./internal
+COPY cmd ./cmd
+COPY go.sum go.sum
+COPY go.mod go.mod
+
 
 # Build our binary.
 RUN go build -o /gotenberg/gotenberg -ldflags "-X main.version=${VERSION}" cmd/gotenberg/main.go
@@ -37,6 +41,8 @@ LABEL authors="Julien Neuhart <j.neuhart@thecodingmachine.com>"
 
 COPY --from=golang /gotenberg/gotenberg /usr/local/bin/
 
+ENV PM2_HOME=/gotenberg/.pm2
+USER gotenberg
 WORKDIR /gotenberg
 
 EXPOSE 3000

cmd/gotenberg/main.go

@@ -25,6 +25,7 @@ const (
 	disableGoogleChromeEnvVar       = "DISABLE_GOOGLE_CHROME"
 	disableUnoconvEnvVar            = "DISABLE_UNOCONV"
 	disableHealthcheckLoggingEnvVar = "DISABLE_HEALTHCHECK_LOGGING"
+	debugProcessStartup             = "DEBUG_PROCESS_STARTUP"
 )
 
 func mustParseEnvVar() *api.Options {
@@ -49,37 +50,31 @@ func mustParseEnvVar() *api.Options {
 		}
 		opts.DefaultListenPort = v
 	}
-	if v, ok := os.LookupEnv(disableGoogleChromeEnvVar); ok {
+	//checkBoolEnv is a convenience function for reading an env var with a bool value where `1` is true and `0` is false
-		if v != "1" && v != "0" {
+	checkBoolEnv := func(name string) bool {
-			notify.ErrPrint(fmt.Errorf("%s: wrong value: want \"0\" or \"1\" got %v", disableGoogleChromeEnvVar, v))
+		if v, ok := os.LookupEnv(name); ok {
-			os.Exit(1)
+			if v != "1" && v != "0" {
+				notify.ErrPrint(fmt.Errorf("%s: wrong value: want \"0\" or \"1\" got %v", name, v))
+				os.Exit(1)
+			}
+			return v == "1"
 		}
-		opts.EnableChromeEndpoints = v != "1"
+		return false
-	}
-	if v, ok := os.LookupEnv(disableUnoconvEnvVar); ok {
-		if v != "1" && v != "0" {
-			notify.ErrPrint(fmt.Errorf("%s: wrong value: want \"0\" or \"1\" got %v", disableUnoconvEnvVar, v))
-			os.Exit(1)
-		}
-		opts.EnableUnoconvEndpoints = v != "1"
-	}
-	if v, ok := os.LookupEnv(disableHealthcheckLoggingEnvVar); ok {
-		if v != "1" && v != "0" {
-			notify.ErrPrint(fmt.Errorf("%s: wrong value: want \"0\" or \"1\" got %v", disableHealthcheckLoggingEnvVar, v))
-			os.Exit(1)
-		}
-		opts.EnableHealthcheckLogging = v != "1"
 	}
+	opts.EnableChromeEndpoints = !checkBoolEnv(disableGoogleChromeEnvVar)
+	opts.EnableUnoconvEndpoints = !checkBoolEnv(disableUnoconvEnvVar)
+	opts.EnableHealthcheckLogging = !checkBoolEnv(disableHealthcheckLoggingEnvVar)
+	opts.DebugProcessStartup = checkBoolEnv(debugProcessStartup)
 	return opts
 }
 
 func mustStartProcesses(opts *api.Options) []pm2.Process {
 	var processes []pm2.Process
 	if opts.EnableChromeEndpoints {
-		processes = append(processes, pm2.NewChrome())
+		processes = append(processes, pm2.NewChrome(opts.DebugProcessStartup))
 	}
 	if opts.EnableUnoconvEndpoints {
-		processes = append(processes, pm2.NewUnoconv())
+		processes = append(processes, pm2.NewUnoconv(opts.DebugProcessStartup))
 	}
 	for _, p := range processes {
 		notify.Printf("starting %s with PM2...", p.Fullname())

docs/index.html

@@ -139,6 +139,8 @@
 <pre class="chroma">$ docker run --rm -p <span class="m">3000</span>:3000 thecodingmachine/gotenberg:5
 </pre>
 
+<p>All processes in the docker container runs as a dedicated non-root user called <code>gotenberg</code> with user id <code>1001</code> from the working dir <code>/gotenberg</code>.</p>
+
 <blockquote>
 <p>The API will be available at <a href="http://localhost:3000">http://localhost:3000</a>.</p>
 </blockquote>
@@ -176,6 +178,13 @@ Otherwise the API will not be able to launch Google Chrome and LibreOffice (unoc
 <p>The more resources are granted, the quicker will be the conversions.</p>
 </blockquote>
 
+<p>Also, in the deployment spec of the pod, specify the uid <code>1001</code> of the user <code>gotenberg</code>:</p>
+
+<pre class="chroma">        securityContext:
+          privileged: false
+          runAsUser: 1001
+</pre>
+
 <p>In the following examples, we will assume your
 Gotenberg API is available at <a href="http://localhost:3000">http://localhost:3000</a>.</p>
 
@@ -222,7 +231,7 @@ Gotenberg API is available at <a href="http://localhost:3000">http://localhost:3
 
 <p>In order to save some resources, the Gotenberg image accepts the environment variable <code>DISABLE_GOOGLE_CHROME</code>.</p>
 
-<p>It takes the strings <code>&#34;0&#34;</code> or <code>&#34;1&#34;</code> as value.</p>
+<p>It takes the strings <code>&#34;0&#34;</code> or <code>&#34;1&#34;</code> as value where <code>1</code> means <code>true</code></p>
 
 <blockquote>
 <p>If Google Chrome is disabled, the following conversions will <strong>not</strong> be available anymore:
@@ -263,7 +272,7 @@ See the <a href="#timeout">timeout section</a>.</p>
 
 <p>You may turn off this logging so as to avoid unnecessary entries in your logs with the environment variable <code>DISABLE_HEALTHCHECK_LOGGING</code>.</p>
 
-<p>This environment variable operates in the same manner as the <code>DISABLE_GOOGLE_CHROME</code> and <code>DISABLE_UNOCONV</code> variables operate in that it accepts the strings <code>&#34;0&#34;</code> or <code>&#34;1&#34;</code> as values.</p>
+<p>This environment variable operates in the same manner as the <code>DISABLE_GOOGLE_CHROME</code> and <code>DISABLE_UNOCONV</code> variables operate in that it accepts the strings <code>&#34;0&#34;</code> or <code>&#34;1&#34;</code> as values, where <code>1</code> is enabled.</p>
 
 <h2 class="Heading"><a class="Anchor" aria-hidden="true" id="environment_variables.default_listen_port" href="#environment_variables.default_listen_port">
 	<svg xmlns="http://www.w3.org/2000/svg" aria-hidden="true" width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="feather feather-link"><path d="M10 13a5 5 0 0 0 7.54.54l3-3a5 5 0 0 0-7.07-7.07l-1.72 1.71"></path><path d="M14 11a5 5 0 0 0-7.54-.54l-3 3a5 5 0 0 0 7.07 7.07l1.71-1.71"></path></svg>
@@ -275,6 +284,16 @@ See the <a href="#timeout">timeout section</a>.</p>
 
 <p>This environment variable accepts any string that can be turned into a port number (e.g., the string <code>&#34;0&#34;</code> up to the string <code>&#34;65535&#34;</code>).</p>
 
+<h2 class="Heading"><a class="Anchor" aria-hidden="true" id="environment_variables.debug_logging_of_process_startup" href="#environment_variables.debug_logging_of_process_startup">
+	<svg xmlns="http://www.w3.org/2000/svg" aria-hidden="true" width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="feather feather-link"><path d="M10 13a5 5 0 0 0 7.54.54l3-3a5 5 0 0 0-7.07-7.07l-1.72 1.71"></path><path d="M14 11a5 5 0 0 0-7.54-.54l-3 3a5 5 0 0 0 7.07 7.07l1.71-1.71"></path></svg>
+</a>Debug logging of process startup</h2>
+
+<p>By default, stdout and stderr messages from the started processes are disabled.</p>
+
+<p>You may enable some debug logging from starting the process by setting the environment variable <code>DEBUG_PROCESS_STARTUP</code>.</p>
+
+<p>This environment variable operates in the same manner as the <code>DISABLE_GOOGLE_CHROME</code> and <code>DISABLE_UNOCONV</code> variables operate in that it accepts the strings <code>&#34;0&#34;</code> or <code>&#34;1&#34;</code> as values, where <code>1</code> means <code>true</code>.</p>
+
               </div>
             
               <div class="Page" id="html">

internal/app/api/api.go

@@ -1,8 +1,6 @@
 package api
 
-import (
+import "github.com/labstack/echo/v4"
-	"github.com/labstack/echo/v4"
-)
 
 const pingEndpoint = "/ping"
 
@@ -14,6 +12,7 @@ type Options struct {
 	EnableChromeEndpoints    bool
 	EnableUnoconvEndpoints   bool
 	EnableHealthcheckLogging bool
+	DebugProcessStartup      bool
 }
 
 // DefaultOptions returns default options.

internal/pkg/pm2/chrome.go

@@ -7,15 +7,19 @@ import (
 	"github.com/mafredri/cdp/devtool"
 )
 
+const (
+	warmupTime = 10 * time.Second
+)
+
 type chrome struct {
 	manager *processManager
 }
 
 // NewChrome retruns a Google Chrome
 // headless process.
-func NewChrome() Process {
+func NewChrome(debug bool) Process {
 	return &chrome{
-		manager: &processManager{},
+		manager: &processManager{verbose: debug},
 	}
 }
 
@@ -58,12 +62,20 @@ func (p *chrome) viable() bool {
 	// check if Google Chrome is correctly running.
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
-	_, err := devtool.New("http://localhost:9222").Version(ctx)
+	p.manager.notifyf(`%v: checking Chrome liveness via debug version endpoint 
-	return err == nil
+'http://localhost:9222/json/version'`, p.name())
+	v, err := devtool.New("http://localhost:9222").Version(ctx)
+	if err != nil {
+		p.manager.notifyf("%s: %s version endpoint returned error: %v", p.name(), p.Fullname(), err)
+		return false
+	}
+	p.manager.notifyf("%s: %s returned version info: %+v", p.name(), p.Fullname(), *v)
+	return true
 }
 
 func (p *chrome) warmup() {
-	time.Sleep(5 * time.Second)
+	p.manager.notifyf("%s: allowing %s %v to startup", p.name(), p.Fullname(), warmupTime)
+	time.Sleep(warmupTime)
 }
 
 // Compile-time checks to ensure type implements desired interfaces.

internal/pkg/pm2/chrome_test.go

@@ -7,13 +7,13 @@ import (
 )
 
 func TestChromeStart(t *testing.T) {
-	p := NewChrome()
+	p := NewChrome(false)
 	err := p.Start()
 	require.Nil(t, err)
 }
 
 func TestChromeShutdown(t *testing.T) {
-	p := NewChrome()
+	p := NewChrome(false)
 	err := p.Shutdown()
 	require.Nil(t, err)
 }

internal/pkg/pm2/pm2.go

@@ -1,8 +1,14 @@
 package pm2
 
 import (
+	"bufio"
 	"fmt"
+	"io"
 	"os/exec"
+	"strings"
+	"time"
+
+	"github.com/thecodingmachine/gotenberg/internal/pkg/notify"
 )
 
 const (
@@ -25,6 +31,7 @@ type Process interface {
 
 type processManager struct {
 	heuristicState int32
+	verbose        bool
 }
 
 func (m *processManager) start(p Process) error {
@@ -76,8 +83,43 @@ func (m *processManager) pm2(p Process, cmdName string) error {
 		"pm2",
 		cmdArgs...,
 	)
+	m.notifyf("executing command '%v'", strings.Join(cmd.Args, " "))
+	if m.verbose {
+		processStdErr, err := cmd.StderrPipe()
+		if err != nil {
+			return fmt.Errorf("failed getting stderr from '%s': %v", p.Fullname(), err)
+		}
+		processStdOut, err := cmd.StdoutPipe()
+		if err != nil {
+			return fmt.Errorf("failed getting stdout from '%s': %v", p.Fullname(), err)
+		}
+		readFromPipe := func(name string, reader io.ReadCloser) {
+			r := bufio.NewReader(reader)
+			defer reader.Close() // nolint: errcheck
+			for {
+				line, _, err := r.ReadLine()
+				if err != nil {
+					if err != io.EOF {
+						m.notifyf("error reading from %v for process '%v'", name, p.Fullname())
+					}
+					break
+				}
+				if len(line) != 0 {
+					m.notifyf("%v %v: %s", p.name(), name, string(line))
+				}
+			}
+		}
+		go readFromPipe("stdout", processStdOut)
+		go readFromPipe("stderr", processStdErr)
+	}
 	if err := cmd.Start(); err != nil {
 		return fmt.Errorf("%s %s with PM2: %v", cmdName, p.Fullname(), err)
 	}
 	return nil
 }
+
+func (m *processManager) notifyf(format string, args ...interface{}) {
+	if m.verbose {
+		notify.Printf(fmt.Sprintf("%v: %s", time.Now().Format(time.RFC3339), format), args...)
+	}
+}

internal/pkg/pm2/unoconv.go

@@ -6,9 +6,9 @@ type unoconv struct {
 
 // NewUnoconv retruns a unoconv listener
 // process.
-func NewUnoconv() Process {
+func NewUnoconv(debug bool) Process {
 	return &unoconv{
-		manager: &processManager{},
+		manager: &processManager{verbose: debug},
 	}
 }
 

internal/pkg/pm2/unoconv_test.go

@@ -7,13 +7,13 @@ import (
 )
 
 func TestUnoconvStart(t *testing.T) {
-	p := NewUnoconv()
+	p := NewUnoconv(false)
 	err := p.Start()
 	require.Nil(t, err)
 }
 
 func TestUnoconvShutdown(t *testing.T) {
-	p := NewUnoconv()
+	p := NewUnoconv(false)
 	err := p.Shutdown()
 	require.Nil(t, err)
 }