X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/b49229f98012d7c08ce02b8d28dbcc165c8a6c53..40f551004ab4e5f1d8ab02ddb55dca225ee8f6ac:/services/keep-balance/main.go

diff --git a/services/keep-balance/main.go b/services/keep-balance/main.go
index 6e89df9a55..37ed06369b 100644
--- a/services/keep-balance/main.go
+++ b/services/keep-balance/main.go
@@ -9,12 +9,17 @@ import (
 	"flag"
 	"fmt"
 	"io"
+	"net/http"
+	_ "net/http/pprof"
 	"os"
 
 	"git.arvados.org/arvados.git/lib/config"
 	"git.arvados.org/arvados.git/lib/service"
 	"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/jmoiron/sqlx"
+	_ "github.com/lib/pq"
 	"github.com/prometheus/client_golang/prometheus"
 	"github.com/sirupsen/logrus"
 )
@@ -27,21 +32,42 @@ func runCommand(prog string, args []string, stdin io.Reader, stdout, stderr io.W
 	logger := ctxlog.FromContext(context.Background())
 
 	var options RunOptions
-	flags := flag.NewFlagSet(prog, flag.ExitOnError)
+	flags := flag.NewFlagSet(prog, flag.ContinueOnError)
 	flags.BoolVar(&options.Once, "once", false,
 		"balance once and then exit")
 	flags.BoolVar(&options.CommitPulls, "commit-pulls", false,
 		"send pull requests (make more replicas of blocks that are underreplicated or are not in optimal rendezvous probe order)")
 	flags.BoolVar(&options.CommitTrash, "commit-trash", false,
 		"send trash requests (delete unreferenced old blocks, and excess replicas of overreplicated blocks)")
-	flags.Bool("version", false, "Write version information to stdout and exit 0")
+	flags.BoolVar(&options.CommitConfirmedFields, "commit-confirmed-fields", true,
+		"update collection fields (replicas_confirmed, storage_classes_confirmed, etc.)")
 	dumpFlag := flags.Bool("dump", false, "dump details for each block to stdout")
+	pprofAddr := flags.String("pprof", "", "serve Go profile data at `[addr]:port`")
+	// "show version" is implemented by service.Command, so we
+	// don't need the var here -- we just need the -version flag
+	// to pass flags.Parse().
+	flags.Bool("version", false, "Write version information to stdout and exit 0")
+
+	if *pprofAddr != "" {
+		go func() {
+			logrus.Println(http.ListenAndServe(*pprofAddr, nil))
+		}()
+	}
 
 	loader := config.NewLoader(os.Stdin, logger)
 	loader.SetupFlags(flags)
 
 	munged := loader.MungeLegacyConfigArgs(logger, args, "-legacy-keepbalance-config")
-	flags.Parse(munged)
+	err := flags.Parse(munged)
+	if err == flag.ErrHelp {
+		return 0
+	} else if err != nil {
+		logger.Errorf("error parsing command line flags: %s", err)
+		return 2
+	} else if flags.NArg() != 0 {
+		logger.Errorf("unrecognized command line arguments: %v", flags.Args())
+		return 2
+	}
 
 	if *dumpFlag {
 		dumper := logrus.New()
@@ -50,11 +76,19 @@ func runCommand(prog string, args []string, stdin io.Reader, stdout, stderr io.W
 		options.Dumper = dumper
 	}
 
-	// Only pass along the version flag, which gets handled in RunCommand
+	// Drop our custom args that would be rejected by the generic
+	// service.Command
 	args = nil
+	dropFlag := map[string]bool{
+		"once":                    true,
+		"commit-pulls":            true,
+		"commit-trash":            true,
+		"commit-confirmed-fields": true,
+		"dump":                    true,
+	}
 	flags.Visit(func(f *flag.Flag) {
-		if f.Name == "version" {
-			args = append(args, "-"+f.Name, f.Value.String())
+		if !dropFlag[f.Name] {
+			args = append(args, "-"+f.Name+"="+f.Value.String())
 		}
 	})
 
@@ -70,6 +104,18 @@ func runCommand(prog string, args []string, stdin io.Reader, stdout, stderr io.W
 				return service.ErrorHandler(ctx, cluster, fmt.Errorf("error initializing client from cluster config: %s", err))
 			}
 
+			db, err := sqlx.Open("postgres", cluster.PostgreSQL.Connection.String())
+			if err != nil {
+				return service.ErrorHandler(ctx, cluster, fmt.Errorf("postgresql connection failed: %s", err))
+			}
+			if p := cluster.PostgreSQL.ConnectionPool; p > 0 {
+				db.SetMaxOpenConns(p)
+			}
+			err = db.Ping()
+			if err != nil {
+				return service.ErrorHandler(ctx, cluster, fmt.Errorf("postgresql connection succeeded but ping failed: %s", err))
+			}
+
 			if options.Logger == nil {
 				options.Logger = ctxlog.FromContext(ctx)
 			}
@@ -81,6 +127,12 @@ func runCommand(prog string, args []string, stdin io.Reader, stdout, stderr io.W
 				Metrics:    newMetrics(registry),
 				Logger:     options.Logger,
 				Dumper:     options.Dumper,
+				DB:         db,
+			}
+			srv.Handler = &health.Handler{
+				Token:  cluster.ManagementToken,
+				Prefix: "/_health/",
+				Routes: health.Routes{"ping": srv.CheckHealth},
 			}
 
 			go srv.run()