Fix Dockerfile

Fixes #22
Credit to @pascaldulieu

Dockerfile

@@ -0,0 +1,12 @@
+FROM golang:alpine as builder
+ADD . /go/src/github.com/terorie/od-database-crawler
+RUN apk add git \
+ && go get -d -v github.com/terorie/od-database-crawler \
+ && CGO_ENABLED=0 go install -a -installsuffix cgo github.com/terorie/od-database-crawler
+
+FROM scratch
+COPY --from=builder /etc/ssl/certs/ca-certificates.crt /etc/ssl/certs/
+COPY --from=builder /go/bin/od-database-crawler /bin/
+WORKDIR /oddb
+VOLUME [ "/oddb" ]
+CMD ["/bin/od-database-crawler", "server"]

README.md

@@ -9,16 +9,46 @@
 
 https://od-db.the-eye.eu/
 
-#### Usage
+## Usage
+
+### Deploys
 
  1. With Config File (if `config.yml` found in working dir)
     - Download [default config](https://github.com/terorie/od-database-crawler/blob/master/config.yml)
     - Set `server.url` and `server.token`
     - Start with `./od-database-crawler server --config <file>`
- 
+
  2. With Flags or env
     - Override config file if it exists
     - `--help` for list of flags
     - Every flag is available as an environment variable:
       `--server.crawl_stats` ➡️ `OD_SERVER_CRAWL_STATS`
     - Start with `./od-database-crawler server <flags>`
+
+ 3. With Docker
+    ```bash
+    docker run \
+        -e OD_SERVER_URL=xxx \
+        -e OD_SERVER_TOKEN=xxx \
+        terorie/od-database-crawler
+    ```
+
+### Flag reference
+
+Here are the most important config flags. For more fine control, take a look at `/config.yml`.
+
+| Flag/Environment                                        | Description                                                  | Example                             |
+| ------------------------------------------------------- | ------------------------------------------------------------ | ----------------------------------- |
+| `server.url`<br />`OD_SERVER_URL`                       | OD-DB Server URL                                             | `https://od-db.mine.the-eye.eu/api` |
+| `server.token`<br />`OD_SERVER_TOKEN`                   | OD-DB Server Access Token                                    | _Ask Hexa **TM**_                   |
+| `server.recheck`<br />`OD_SERVER_RECHECK`               | Job Fetching Interval                                        | `3s`                                |
+| `output.crawl_stats`<br />`OD_OUTPUT_CRAWL_STATS`       | Crawl Stats Logging Interval (0 = disabled)                  | `500ms`                             |
+| `output.resource_stats`<br />`OD_OUTPUT_RESORUCE_STATS` | Resource Stats Logging Interval (0 = disabled)               | `8s`                                |
+| `output.log`<br />`OD_OUTPUT_LOG`                       | Log File (none = disabled)                                   | `crawler.log`                       |
+| `crawl.tasks`<br />`OD_CRAWL_TASKS`                     | Max number of sites to crawl concurrently                    | `500`                               |
+| `crawl.connections`<br />`OD_CRAWL_CONNECTIONS`         | HTTP connections per site                                    | `1`                                 |
+| `crawl.retries`<br />`OD_CRAWL_RETRIES`                 | How often to retry after a temporary failure (e.g. `HTTP 429` or timeouts) | `5`                                 |
+| `crawl.dial_timeout`<br />`OD_CRAWL_DIAL_TIMEOUT`       | TCP Connect timeout                                          | `5s`                                |
+| `crawl.timeout`<br />`OD_CRAWL_TIMEOUT`                 | HTTP request timeout                                         | `20s`                               |
+| `crawl.user-agent`<br />`OD_CRAWL_USER_AGENT`           | HTTP Crawler User-Agent                                      | `googlebot/1.2.3`                   |
+| `crawl.job_buffer`<br />`OD_CRAWL_JOB_BUFFER`           | Number of URLs to keep in memory/cache, per job. The rest is offloaded to disk. Decrease this value if the crawler uses too much RAM. (0 = Disable Cache, -1 = Only use Cache) | `5000`                              |

config.go

@@ -4,10 +4,10 @@ import (
 	"bufio"
 	"fmt"
 	"github.com/sirupsen/logrus"
+	"github.com/spf13/pflag"
 	"github.com/spf13/viper"
 	"io"
 	"os"
-	"path/filepath"
 	"strings"
 	"time"
 )
@@ -27,6 +27,8 @@ var config struct {
 	JobBufferSize int
 }
 
+var onlineMode bool
+
 const (
 	ConfServerUrl  = "server.url"
 	ConfToken      = "server.token"
@@ -55,8 +57,56 @@ const (
 func prepareConfig() {
 	pf := rootCmd.PersistentFlags()
 
-	bind := func(s string) {
-		if err := viper.BindPFlag(s, pf.Lookup(s)); err != nil {
+	pf.SortFlags = false
+	pf.StringVar(&configFile, "config", "", "Config file")
+	configFile = os.Getenv("OD_CONFIG")
+
+	pf.String(ConfServerUrl, "http://od-db.the-eye.eu/api", "OD-DB server URL")
+
+	pf.String(ConfToken, "", "OD-DB access token (env OD_SERVER_TOKEN)")
+
+	pf.Duration(ConfServerTimeout, 60 * time.Second, "OD-DB request timeout")
+
+	pf.Duration(ConfRecheck, 1 * time.Second, "OD-DB: Poll interval for new jobs")
+
+	pf.Duration(ConfCooldown, 30 * time.Second, "OD-DB: Time to wait after a server-side error")
+
+	pf.String(ConfChunkSize, "1 MB", "OD-DB: Result upload chunk size")
+
+	pf.Uint(ConfUploadRetries, 10, "OD-DB: Max upload retries")
+
+	pf.Duration(ConfUploadRetryInterval, 30 * time.Second, "OD-DB: Time to wait between upload retries")
+
+	pf.Uint(ConfTasks, 100, "Crawler: Max concurrent tasks")
+
+	pf.Uint(ConfWorkers, 4, "Crawler: Connections per server")
+
+	pf.Uint(ConfRetries, 5, "Crawler: Request retries")
+
+	pf.Duration(ConfDialTimeout, 10 * time.Second, "Crawler: Handshake timeout")
+
+	pf.Duration(ConfTimeout, 30 * time.Second, "Crawler: Request timeout")
+
+	pf.String(ConfUserAgent, "Mozilla/5.0 (X11; od-database-crawler) Gecko/20100101 Firefox/52.0", "Crawler: User-Agent")
+
+	pf.Uint(ConfJobBufferSize, 5000, "Crawler: Task queue cache size")
+
+	pf.Duration(ConfCrawlStats, time.Second, "Log: Crawl stats interval")
+
+	pf.Duration(ConfAllocStats, 10 * time.Second, "Log: Resource stats interval")
+
+	pf.Bool(ConfVerbose, false, "Log: Print every listed dir")
+
+	pf.Bool(ConfPrintHTTP, false, "Log: Print HTTP client errors")
+
+	pf.String(ConfLogFile, "crawler.log", "Log file")
+
+	// Bind all flags to Viper
+	pf.VisitAll(func(flag *pflag.Flag) {
+		s := flag.Name
+		s = strings.TrimLeft(s, "-")
+
+		if err := viper.BindPFlag(s, flag); err != nil {
 			panic(err)
 		}
 		var envKey string
@@ -66,101 +116,45 @@ func prepareConfig() {
 		if err := viper.BindEnv(s, envKey); err != nil {
 			panic(err)
 		}
-	}
-
-	pf.SortFlags = false
-	pf.StringVar(&configFile, "config", "", "Config file")
-	configFile = os.Getenv("OD_CONFIG")
-
-	pf.String(ConfServerUrl, "http://od-db.the-eye.eu/api", "OD-DB server URL")
-	bind(ConfServerUrl)
-
-	pf.String(ConfToken, "", "OD-DB access token (env OD_SERVER_TOKEN)")
-	bind(ConfToken)
-
-	pf.Duration(ConfServerTimeout, 60 * time.Second, "OD-DB request timeout")
-	bind(ConfServerTimeout)
-
-	pf.Duration(ConfRecheck, 1 * time.Second, "OD-DB: Poll interval for new jobs")
-	bind(ConfRecheck)
-
-	pf.Duration(ConfCooldown, 30 * time.Second, "OD-DB: Time to wait after a server-side error")
-	bind(ConfCooldown)
-
-	pf.String(ConfChunkSize, "1 MB", "OD-DB: Result upload chunk size")
-	bind(ConfChunkSize)
-
-	pf.Uint(ConfUploadRetries, 10, "OD-DB: Max upload retries")
-	bind(ConfUploadRetries)
-
-	pf.Duration(ConfUploadRetryInterval, 30 * time.Second, "OD-DB: Time to wait between upload retries")
-	bind(ConfUploadRetryInterval)
-
-	pf.Uint(ConfTasks, 100, "Crawler: Max concurrent tasks")
-	bind(ConfTasks)
-
-	pf.Uint(ConfWorkers, 4, "Crawler: Connections per server")
-	bind(ConfWorkers)
-
-	pf.Uint(ConfRetries, 5, "Crawler: Request retries")
-	bind(ConfRetries)
-
-	pf.Duration(ConfDialTimeout, 10 * time.Second, "Crawler: Handshake timeout")
-	bind(ConfDialTimeout)
-
-	pf.Duration(ConfTimeout, 30 * time.Second, "Crawler: Request timeout")
-	bind(ConfTimeout)
-
-	pf.String(ConfUserAgent, "Mozilla/5.0 (X11; od-database-crawler) Gecko/20100101 Firefox/52.0", "Crawler: User-Agent")
-	bind(ConfUserAgent)
-
-	pf.Uint(ConfJobBufferSize, 5000, "Crawler: Task queue cache size")
-	bind(ConfJobBufferSize)
-
-	pf.Duration(ConfCrawlStats, time.Second, "Log: Crawl stats interval")
-	bind(ConfCrawlStats)
-
-	pf.Duration(ConfAllocStats, 10 * time.Second, "Log: Resource stats interval")
-	bind(ConfAllocStats)
-
-	pf.Bool(ConfVerbose, false, "Log: Print every listed dir")
-	bind(ConfVerbose)
-
-	pf.Bool(ConfPrintHTTP, false, "Log: Print HTTP client errors")
-	bind(ConfPrintHTTP)
-
-	pf.String(ConfLogFile, "crawler.log", "Log file")
-	bind(ConfLogFile)
+	})
 }
 
 func readConfig() {
 	// If config.yml in working dir, use it
-	if _, err := os.Stat("config.yml"); err == nil {
-		configFile = "config.yml"
+	if configFile == "" {
+		_, err := os.Stat("config.yml")
+		if err == nil {
+			configFile = "config.yml"
+		}
 	}
 
 	if configFile != "" {
-		var err error
-		confPath, err := filepath.Abs(configFile)
-		if err != nil { panic(err) }
+		confF, err := os.Open(configFile)
+		if err != nil {
+			fmt.Fprintln(os.Stderr, err)
+			os.Exit(1)
+		}
+		defer confF.Close()
 
-		viper.SetConfigFile(confPath)
-		err = viper.ReadInConfig()
+		viper.SetConfigType("yml")
+		err = viper.ReadConfig(confF)
 		if err != nil {
 			fmt.Fprintln(os.Stderr, err)
 			os.Exit(1)
 		}
 	}
 
-	config.ServerUrl = viper.GetString(ConfServerUrl)
-	if config.ServerUrl == "" {
-		configMissing(ConfServerUrl)
-	}
-	config.ServerUrl = strings.TrimRight(config.ServerUrl, "/")
+	if onlineMode {
+		config.ServerUrl = viper.GetString(ConfServerUrl)
+		if config.ServerUrl == "" {
+			configMissing(ConfServerUrl)
+		}
+		config.ServerUrl = strings.TrimRight(config.ServerUrl, "/")
 
-	config.Token = viper.GetString(ConfToken)
-	if config.Token == "" {
-		configMissing(ConfToken)
+		config.Token = viper.GetString(ConfToken)
+		if config.Token == "" {
+			configMissing(ConfToken)
+		}
 	}
 
 	config.ServerTimeout = viper.GetDuration(ConfServerTimeout)

config.yml

@@ -53,7 +53,7 @@ crawl:
   # Please be careful with this setting!
   # The crawler fires fast and more than
   # ten connections can overwhelm a server.
-  connections: 4
+  connections: 1
 
   # How often to retry getting data
   # from the site before giving up

errors.go

@@ -3,6 +3,8 @@ package main
 import (
 	"errors"
 	"fmt"
+	"github.com/valyala/fasthttp"
+	"net"
 )
 
 var ErrRateLimit = errors.New("too many requests")
@@ -15,3 +17,29 @@ type HttpError struct {
 func (e HttpError) Error() string {
 	return fmt.Sprintf("http status %d", e.code)
 }
+
+func shouldRetry(err error) bool {
+	// HTTP errors
+	if httpErr, ok := err.(*HttpError); ok {
+		switch httpErr.code {
+		case fasthttp.StatusTooManyRequests:
+			return true
+		default:
+			// Don't retry HTTP error codes
+			return false
+		}
+	}
+
+	if dnsError, ok := err.(*net.DNSError); ok {
+		// Don't retry permanent DNS errors
+		return dnsError.IsTemporary
+	}
+
+	if netErr, ok := err.(*net.OpError); ok {
+		// Don't retry permanent network errors
+		return netErr.Temporary()
+	}
+
+	// Retry by default
+	return true
+}

go.mod

@@ -1,4 +1,4 @@
-module github.com/syndtr/od-database-crawler
+module github.com/terorie/od-database-crawler
 
 require (
 	github.com/beeker1121/goque v2.0.1+incompatible
@@ -7,7 +7,6 @@ require (
 	github.com/spf13/cobra v0.0.3
 	github.com/spf13/viper v1.3.1
 	github.com/syndtr/goleveldb v0.0.0-20181128100959-b001fa50d6b2 // indirect
-	github.com/terorie/od-database-crawler v1.1.1
 	github.com/valyala/fasthttp v1.1.0
 	golang.org/x/crypto v0.0.0-20190131182504-b8fe1690c613
 	golang.org/x/net v0.0.0-20180911220305-26e67e76b6c3

go.sum

@@ -41,8 +41,6 @@ github.com/stretchr/objx v0.1.1/go.mod h1:HFkY916IF+rwdDfMAkV7OtwuqBVzrE8GR6GFx+
 github.com/stretchr/testify v1.2.2/go.mod h1:a8OnRcib4nhh0OaRAV+Yts87kKdq0PP7pXfy6kDkUVs=
 github.com/syndtr/goleveldb v0.0.0-20181128100959-b001fa50d6b2 h1:GnOzE5fEFN3b2zDhJJABEofdb51uMRNb8eqIVtdducs=
 github.com/syndtr/goleveldb v0.0.0-20181128100959-b001fa50d6b2/go.mod h1:Z4AUp2Km+PwemOoO/VB5AOx9XSsIItzFjoJlOSiYmn0=
-github.com/terorie/od-database-crawler v1.1.1 h1:Ca+ZqbZX3rVWBR8SDRzvroyxjBtUs75MQXZ9YG0gqGo=
-github.com/terorie/od-database-crawler v1.1.1/go.mod h1:vVJ7pLkudrlUNp9qu24JCzQ8N6mFsrOmX1tPXr155DQ=
 github.com/ugorji/go/codec v0.0.0-20181204163529-d75b2dcb6bc8/go.mod h1:VFNgLljTbGfSG7qAOspJ7OScBnGdDN/yBr0sguwnwf0=
 github.com/valyala/bytebufferpool v1.0.0 h1:GqA5TC/0021Y/b9FG4Oi9Mr3q7XYx6KllzawFIhcdPw=
 github.com/valyala/bytebufferpool v1.0.0/go.mod h1:6bBcMArwyJ5K/AmCkWv1jt77kVWyCJ6HpOuEn7z0Csc=

main.go

@@ -8,6 +8,7 @@ import (
 	"github.com/spf13/viper"
 	"github.com/terorie/od-database-crawler/fasturl"
 	"os"
+	"os/signal"
 	"strings"
 	"sync/atomic"
 	"time"
@@ -17,7 +18,7 @@ var configFile string
 
 var rootCmd = cobra.Command {
 	Use: "od-database-crawler",
-	Version: "1.2.0",
+	Version: "1.2.2",
 	Short: "OD-Database Go crawler",
 	Long: helpText,
 	PersistentPreRunE: preRun,
@@ -61,8 +62,6 @@ func preRun(cmd *cobra.Command, args []string) error {
 	if err := os.MkdirAll("queue", 0755);
 		err != nil { panic(err) }
 
-	readConfig()
-
 	return nil
 }
 
@@ -75,25 +74,31 @@ func main() {
 }
 
 func cmdBase(_ *cobra.Command, _ []string) {
-	// TODO Graceful shutdown
-	appCtx := context.Background()
-	forceCtx := context.Background()
+	onlineMode = true
+	readConfig()
+
+	appCtx, soft := context.WithCancel(context.Background())
+	forceCtx, hard := context.WithCancel(context.Background())
+	go hardShutdown(forceCtx)
+	go listenCtrlC(soft, hard)
 
 	inRemotes := make(chan *OD)
-	go Schedule(forceCtx, inRemotes)
+	go Schedule(appCtx, inRemotes)
 
 	ticker := time.NewTicker(config.Recheck)
 	defer ticker.Stop()
 	for {
 		select {
 		case <-appCtx.Done():
-			return
+			goto shutdown
 		case <-ticker.C:
 			t, err := FetchTask()
 			if err != nil {
 				logrus.WithError(err).
 					Error("Failed to get new task")
-				time.Sleep(viper.GetDuration(ConfCooldown))
+				if !sleep(viper.GetDuration(ConfCooldown), appCtx) {
+					goto shutdown
+				}
 				continue
 			}
 			if t == nil {
@@ -109,13 +114,7 @@ func cmdBase(_ *cobra.Command, _ []string) {
 			if urlErr, ok := err.(*fasturl.Error); ok && urlErr.Err == fasturl.ErrUnknownScheme {
 				// Not an error
 				err = nil
-
-				// Give back task
-				//err2 := CancelTask(t.WebsiteId)
-				//if err2 != nil {
-				//	logrus.Error(err2)
-				//}
-
+				// TODO FTP crawler
 				continue
 			} else if err != nil {
 				logrus.WithError(err).
@@ -126,9 +125,15 @@ func cmdBase(_ *cobra.Command, _ []string) {
 			ScheduleTask(inRemotes, t, &baseUri)
 		}
 	}
+
+shutdown:
+	globalWait.Wait()
 }
 
 func cmdCrawler(_ *cobra.Command, args []string) error {
+	onlineMode = false
+	readConfig()
+
 	arg := args[0]
 	// https://github.com/golang/go/issues/19779
 	if !strings.Contains(arg, "://") {
@@ -161,3 +166,30 @@ func cmdCrawler(_ *cobra.Command, args []string) error {
 
 	return nil
 }
+
+func listenCtrlC(soft, hard context.CancelFunc) {
+	c := make(chan os.Signal)
+	signal.Notify(c, os.Interrupt)
+
+	<-c
+	logrus.Info(">>> Shutting down crawler... <<<")
+	soft()
+
+	<-c
+	logrus.Warning(">>> Force shutdown! <<<")
+	hard()
+}
+
+func hardShutdown(c context.Context) {
+	<-c.Done()
+	os.Exit(1)
+}
+
+func sleep(d time.Duration, c context.Context) bool {
+	select {
+	case <-time.After(d):
+		return true
+	case <-c.Done():
+		return false
+	}
+}

server.go

@@ -17,8 +17,11 @@ import (
 
 var serverClient = http.Client {
 	Timeout: config.ServerTimeout,
+	Transport: new(ServerTripper),
 }
 
+var serverUserAgent = "od-database-crawler/" + rootCmd.Version
+
 func FetchTask() (t *Task, err error) {
 	res, err := serverClient.PostForm(
 		config.ServerUrl + "/task/get",
@@ -37,7 +40,9 @@ func FetchTask() (t *Task, err error) {
 
 	t = new(Task)
 	err = json.NewDecoder(res.Body).Decode(t)
-	if err != nil { return }
+	if _, ok := err.(*json.SyntaxError); ok {
+		return nil, fmt.Errorf("/task/get returned invalid JSON")
+	} else if err != nil { return }
 
 	return
 }
@@ -176,3 +181,10 @@ func CancelTask(websiteId uint64) (err error) {
 
 	return
 }
+
+type ServerTripper struct{}
+
+func (t *ServerTripper) RoundTrip(req *http.Request) (res *http.Response, err error) {
+	req.Header.Set("User-Agent", serverUserAgent)
+	return http.DefaultTransport.RoundTrip(req)
+}

worker.go

@@ -3,7 +3,6 @@ package main
 import (
 	"github.com/beeker1121/goque"
 	"github.com/sirupsen/logrus"
-	"github.com/valyala/fasthttp"
 	"math"
 	"sort"
 	"strings"
@@ -55,14 +54,12 @@ func (w *WorkerContext) step(results chan<- File, job Job) {
 	if err != nil {
 		job.Fails++
 
-		if httpErr, ok := err.(*HttpError); ok {
-			switch httpErr.code {
-			case fasthttp.StatusTooManyRequests:
-				err = ErrRateLimit
-			default:
-				// Don't retry HTTP error codes
-				return
-			}
+		if !shouldRetry(err) {
+			atomic.AddUint64(&totalAborted, 1)
+			logrus.WithField("url", job.UriStr).
+				WithError(err).
+				Error("Giving up after failure")
+			return
 		}
 
 		if job.Fails > config.Retries {