CAS-1379 Basic process management for mon/osd procs

glide.lock

@@ -1,5 +1,5 @@
-hash: 92b0b86293cd8ff7e45e4b92b4d116a38ea8c3a7b92137fd618742f3c67a7b15
-updated: 2016-08-31T09:18:34.23005271-07:00
+hash: e304bcf51126db7c40de2ad0abe6cf24b9d5153ef72b49244c5722f30af4e61d
+updated: 2016-09-08T13:07:22.421779478-07:00
 imports:
 - name: github.com/coreos/etcd
   version: d53923c636e0e4ab7f00cb75681b97a8f11f5a9d
@@ -15,11 +15,12 @@ imports:
 - name: github.com/inconshreveable/mousetrap
   version: 76626ae9c91c4f2a10f34cad8ce83ea42c93bb75
 - name: github.com/quantum/clusterd
-  version: 27cbadad4d31c6769e8bd3d0cd4d7908f1fe5590
+  version: dfc6f8c199d906c4206cce9e9f80a301ad1f22b8
   subpackages:
   - pkg/inventory
   - pkg/orchestrator
   - pkg/util
+  - pkg/proc
 - name: github.com/spf13/cobra
   version: dbb7c2d02e284aa24423e3e6e105969dffe6e4d6
   subpackages:

glide.yaml

@@ -8,10 +8,12 @@ import:
   - client
   - error
 - package: github.com/quantum/clusterd
+  version: dfc6f8c199d906c4206cce9e9f80a301ad1f22b8
   subpackages:
   - pkg/inventory
   - pkg/orchestrator
   - pkg/util
+  - pkg/proc
 - package: github.com/spf13/pflag
 - package: github.com/google/uuid
 - package: github.com/go-ini/ini

pkg/castled/cmd/castled.go

@@ -4,13 +4,14 @@ import (
 	"fmt"
 	"os"
 	"os/signal"
+	"time"
 
 	"github.com/spf13/cobra"
 
 	"github.com/quantum/castle/pkg/castled"
 	"github.com/quantum/castle/pkg/util"
 	"github.com/quantum/clusterd/pkg/orchestrator"
-	clusterutil "github.com/quantum/clusterd/pkg/util"
+	"github.com/quantum/clusterd/pkg/proc"
 )
 
 var (
@@ -56,8 +57,11 @@ func joinCluster(cmd *cobra.Command, args []string) error {
 	}
 
 	services := []*orchestrator.ClusterService{castled.NewCephService(devices, forceFormat)}
-	procMan := &clusterutil.ProcManager{}
-	defer procMan.Shutdown()
+	procMan := &proc.ProcManager{}
+	defer func() {
+		procMan.Shutdown()
+		<-time.After(time.Duration(1) * time.Second)
+	}()
 
 	// start the cluster orchestration services
 	if err := orchestrator.StartJoinCluster(services, procMan, discoveryURL, privateIPv4); err != nil {

pkg/castled/monagent.go

@@ -7,11 +7,13 @@ import (
 	"os"
 	"path"
 	"path/filepath"
+	"regexp"
 	"time"
 
 	ctx "golang.org/x/net/context"
 
 	"github.com/quantum/clusterd/pkg/orchestrator"
+	"github.com/quantum/clusterd/pkg/proc"
 	"github.com/quantum/clusterd/pkg/util"
 )
 
@@ -138,11 +140,14 @@ func (a *monAgent) runMonitor(context *orchestrator.ClusterContext, monitor *Cep
 
 	// start the monitor daemon in the foreground with the given config
 	log.Printf("starting monitor %s", monitor.Name)
+	monNameArg := fmt.Sprintf("--name=mon.%s", monitor.Name)
 	err := context.ProcMan.Start(
 		"mon",
+		regexp.QuoteMeta(monNameArg),
+		proc.ReuseExisting,
 		"--foreground",
-		fmt.Sprintf("--cluster=%v", a.cluster.Name),
-		fmt.Sprintf("--name=mon.%v", monitor.Name),
+		fmt.Sprintf("--cluster=%s", a.cluster.Name),
+		monNameArg,
 		fmt.Sprintf("--mon-data=%s", getMonDataDirPath(monitor.Name)),
 		fmt.Sprintf("--conf=%s", getConfFilePath(getMonRunDirPath(monitor.Name), a.cluster.Name)),
 		fmt.Sprintf("--public-addr=%s", monitor.Endpoint))

pkg/castled/osd.go

@@ -16,6 +16,7 @@ import (
 
 	"github.com/google/uuid"
 	"github.com/quantum/castle/pkg/cephd"
+	"github.com/quantum/clusterd/pkg/proc"
 	"github.com/quantum/clusterd/pkg/util"
 )
 
@@ -400,7 +401,7 @@ func createOSDFileSystem(clusterName string, osdID int, osdUUID uuid.UUID, osdDa
 	}
 
 	// create the OSD file system and journal
-	err := util.RunChildProcess(
+	err := proc.RunChildProcess(
 		"osd",
 		"--mkfs",
 		"--mkkey",

pkg/castled/osdagent.go

@@ -6,6 +6,7 @@ import (
 	"os"
 	"path"
 	"path/filepath"
+	"regexp"
 	"strconv"
 	"strings"
 
@@ -13,6 +14,7 @@ import (
 
 	"github.com/quantum/castle/pkg/cephd"
 	"github.com/quantum/clusterd/pkg/orchestrator"
+	"github.com/quantum/clusterd/pkg/proc"
 	"github.com/quantum/clusterd/pkg/util"
 )
 
@@ -157,8 +159,11 @@ func (a *osdAgent) createOSDs(adminConn *cephd.Conn, context *orchestrator.Clust
 func (a *osdAgent) runOSD(context *orchestrator.ClusterContext, clusterName string, osdID int, osdUUID uuid.UUID, osdDataPath string) error {
 	// start the OSD daemon in the foreground with the given config
 	log.Printf("starting osd %d at %s", osdID, osdDataPath)
+	osdUUIDArg := fmt.Sprintf("--osd-uuid=%s", osdUUID.String())
 	err := context.ProcMan.Start(
 		"osd",
+		regexp.QuoteMeta(osdUUIDArg),
+		proc.ReuseExisting,
 		"--foreground",
 		fmt.Sprintf("--id=%s", strconv.Itoa(osdID)),
 		fmt.Sprintf("--cluster=%s", clusterName),
@@ -166,7 +171,7 @@ func (a *osdAgent) runOSD(context *orchestrator.ClusterContext, clusterName stri
 		fmt.Sprintf("--osd-journal=%s", getOSDJournalPath(osdDataPath)),
 		fmt.Sprintf("--conf=%s", getOSDConfFilePath(osdDataPath, clusterName)),
 		fmt.Sprintf("--keyring=%s", getOSDKeyringPath(osdDataPath)),
-		fmt.Sprintf("--osd-uuid=%s", osdUUID.String()))
+		osdUUIDArg)
 	if err != nil {
 		return fmt.Errorf("failed to start osd %d: %+v", osdID, err)
 	}