mon: set names based on chars instead of integers

Signed-off-by: travisn <tnielsen@redhat.com>

PendingReleaseNotes.md

@@ -12,6 +12,7 @@
 - The minimum version of Kubernetes supported by Rook changed from `1.7` to `1.8`.
 
 ## Breaking Changes
+- Mons are [named consistently](https://github.com/rook/rook/issues/1751) with other daemons with the letters a, b, c, etc.
 
 - The Rook container images are no longer published to quay.io, they are published only to Docker Hub.  All manifests have referenced Docker Hub for multiple releases now, so we do not expect any directly affected users from this change.
 - Rook no longer supports kubernetes `1.7`. Users running Kubernetes `1.7` on their clusters are recommended to upgrade to Kubernetes `1.8` or higher. If you are using `kubeadm`, you can follow this [guide](https://kubernetes.io/docs/tasks/administer-cluster/kubeadm/kubeadm-upgrade-1-8/) to from Kubernetes `1.7` to `1.8`. If you are using `kops` or `kubespray` for managing your Kubernetes cluster, just follow the respective projects' `upgrade` guide.

cmd/rook/ceph/mon.go

@@ -66,7 +66,7 @@ func startMon(cmd *cobra.Command, args []string) error {
 		return fmt.Errorf("missing mon port")
 	}
 
-	if err := compareMonSecret(clusterInfo.MonitorSecret, path.Join(cfg.dataDir, monName)); err != nil {
+	if err := compareMonSecret(clusterInfo.MonitorSecret, mon.GetMonRunDirPath(cfg.dataDir, monName)); err != nil {
 		rook.TerminateFatal(err)
 	}
 

images/ceph-toolbox/toolbox.sh

@@ -28,7 +28,13 @@ KEYRING_FILE="/etc/ceph/keyring"
 # without specifying any arguments
 write_endpoints() {
     endpoints=$(cat ${MON_CONFIG})
-    mon_endpoints=$(echo ${endpoints} | sed 's/rook-ceph-mon[0-9]\+=//g')
+
+    # filter out the mon names
+    mon_endpoints=$(echo ${endpoints} | sed 's/[a-z]\+=//g')
+
+    # filter out the legacy mon names
+    mon_endpoints=$(echo ${mon_endpoints} | sed 's/rook-ceph-mon[0-9]\+=//g')
+
     DATE=$(date)
     echo "$DATE writing mon endpoints to ${CEPH_CONFIG}: ${endpoints}"
     cat <<EOF > ${CEPH_CONFIG}

pkg/daemon/ceph/client/test/mon.go

@@ -26,7 +26,7 @@ func MonInQuorumResponse() string {
 	resp := client.MonStatusResponse{Quorum: []int{0}}
 	resp.MonMap.Mons = []client.MonMapEntry{
 		{
-			Name:    "rook-ceph-mon1",
+			Name:    "rook-ceph-mon-a",
 			Rank:    0,
 			Address: "1.2.3.1",
 		},

pkg/daemon/ceph/mon/config.go

@@ -99,18 +99,22 @@ type GlobalConfig struct {
 }
 
 // get the path of a given monitor's run dir
-func getMonRunDirPath(configDir, monName string) string {
+func GetMonRunDirPath(configDir, monName string) string {
+	if strings.Index(monName, "mon") == -1 {
+		// if the mon name doesn't have "mon" in it, include it in the directory
+		return path.Join(configDir, "mon-"+monName)
+	}
 	return path.Join(configDir, monName)
 }
 
 // get the path of a given monitor's keyring
 func getMonKeyringPath(configDir, monName string) string {
-	return filepath.Join(getMonRunDirPath(configDir, monName), defaultKeyringFile)
+	return filepath.Join(GetMonRunDirPath(configDir, monName), defaultKeyringFile)
 }
 
 // get the path of a given monitor's data dir
 func getMonDataDirPath(configDir, monName string) string {
-	return filepath.Join(getMonRunDirPath(configDir, monName), "data")
+	return filepath.Join(GetMonRunDirPath(configDir, monName), "data")
 }
 
 // get the path of a given monitor's config file
@@ -190,7 +194,7 @@ func GenerateConfigFile(context *clusterd.Context, cluster *ClusterInfo, pathRoo
 	globalConfig *cephConfig, clientSettings map[string]string) (string, error) {
 
 	if pathRoot == "" {
-		pathRoot = getMonRunDirPath(context.ConfigDir, getFirstMonitor(cluster))
+		pathRoot = GetMonRunDirPath(context.ConfigDir, getFirstMonitor(cluster))
 	}
 
 	// create the config directory

pkg/daemon/ceph/mon/daemon.go

@@ -89,7 +89,7 @@ func generateConfigFiles(context *clusterd.Context, config *Config) (string, str
 	}
 
 	// write the config file to disk
-	confFilePath, err := GenerateConnectionConfigFile(context, config.Cluster, getMonRunDirPath(context.ConfigDir, config.Name),
+	confFilePath, err := GenerateConnectionConfigFile(context, config.Cluster, GetMonRunDirPath(context.ConfigDir, config.Name),
 		"admin", getMonKeyringPath(context.ConfigDir, config.Name))
 	if err != nil {
 		return "", "", err
@@ -114,7 +114,7 @@ func startMon(context *clusterd.Context, config *Config, confFilePath, monDataDi
 	logger.Infof("initializing mon")
 
 	// generate the monmap
-	monmapPath, err := generateMonMap(context, config.Cluster, getMonRunDirPath(context.ConfigDir, config.Name))
+	monmapPath, err := generateMonMap(context, config.Cluster, GetMonRunDirPath(context.ConfigDir, config.Name))
 	if err != nil {
 		return err
 	}

pkg/operator/ceph/cluster/mon/health.go

@@ -233,8 +233,8 @@ func (c *Cluster) failoverMon(name string) error {
 	logger.Infof("Failing over monitor %s", name)
 
 	// Start a new monitor
-	m := &monConfig{Name: fmt.Sprintf("%s%d", appName, c.maxMonID+1), Port: int32(mon.DefaultPort)}
-	logger.Infof("starting new mon %s", m.Name)
+	m := newMonConfig(c.maxMonID + 1)
+	logger.Infof("starting new mon: %+v", m)
 
 	// Create the service endpoint
 	serviceIP, err := c.createService(m)
@@ -258,7 +258,7 @@ func (c *Cluster) failoverMon(name string) error {
 	} else {
 		m.PublicIP = serviceIP
 	}
-	c.clusterInfo.Monitors[m.Name] = mon.ToCephMon(m.Name, m.PublicIP, m.Port)
+	c.clusterInfo.Monitors[m.Name] = mon.ToCephMon(m.DaemonName, m.PublicIP, m.Port)
 
 	// Start the pod
 	if err = c.startPods(mConf); err != nil {

pkg/operator/ceph/cluster/mon/health_test.go

@@ -55,30 +55,30 @@ func TestCheckHealth(t *testing.T) {
 	c := New(context, "ns", "", "myversion", cephv1beta1.MonSpec{Count: 3, AllowMultiplePerNode: true},
 		rookalpha.Placement{}, false, v1.ResourceRequirements{}, metav1.OwnerReference{})
 	c.clusterInfo = test.CreateConfigDir(1)
+	logger.Infof("initial mons: %v", c.clusterInfo.Monitors)
 	c.waitForStart = false
 	defer os.RemoveAll(c.context.ConfigDir)
 
-	c.mapping.Node["rook-ceph-mon1"] = &NodeInfo{
+	c.mapping.Node["rook-ceph-mon-f"] = &NodeInfo{
 		Name:    "node0",
 		Address: "",
 	}
 	c.mapping.Port["node0"] = cephmon.DefaultPort
-	c.maxMonID = 10
+	c.maxMonID = 4
 
 	err := c.checkHealth()
 	assert.Nil(t, err)
+	logger.Infof("mons after checkHealth: %v", c.clusterInfo.Monitors)
 
-	err = c.failoverMon("rook-ceph-mon1")
+	err = c.failoverMon("f")
 	assert.Nil(t, err)
 
 	newMons := []string{
-		"rook-ceph-mon11",
-		"rook-ceph-mon12",
-		"rook-ceph-mon13",
+		"rook-ceph-mon-g",
 	}
 	for _, monName := range newMons {
 		_, ok := c.clusterInfo.Monitors[monName]
-		assert.True(t, ok, fmt.Sprintf("mon %s not found in monitor list", monName))
+		assert.True(t, ok, fmt.Sprintf("mon %s not found in monitor list. %v", monName, c.clusterInfo.Monitors))
 	}
 }
 
@@ -102,39 +102,34 @@ func TestCheckHealthNotFound(t *testing.T) {
 	c.waitForStart = false
 	defer os.RemoveAll(c.context.ConfigDir)
 
-	c.mapping.Node["rook-ceph-mon1"] = &NodeInfo{
-		Name:    "node0",
-		Address: "",
+	c.mapping.Node["rook-ceph-mon-a"] = &NodeInfo{
+		Name: "node0",
 	}
-	c.mapping.Node["rook-ceph-mon2"] = &NodeInfo{
-		Name:    "node0",
-		Address: "",
+	c.mapping.Node["rook-ceph-mon-b"] = &NodeInfo{
+		Name: "node0",
 	}
 	c.mapping.Port["node0"] = cephmon.DefaultPort
-	c.maxMonID = 10
+	c.maxMonID = 4
 
 	c.saveMonConfig()
 
 	// Check if the two mons are found in the configmap
 	cm, err := c.context.Clientset.CoreV1().ConfigMaps(c.Namespace).Get(EndpointConfigMapName, metav1.GetOptions{})
 	assert.Nil(t, err)
-	if cm.Data[EndpointDataKey] == "rook-ceph-mon1=1.2.3.1:6790,rook-ceph-mon2=1.2.3.2:6790" {
-		assert.Equal(t, "rook-ceph-mon1=1.2.3.1:6790,rook-ceph-mon2=1.2.3.2:6790", cm.Data[EndpointDataKey])
-	} else {
-		assert.Equal(t, "rook-ceph-mon2=1.2.3.2:6790,rook-ceph-mon1=1.2.3.1:6790", cm.Data[EndpointDataKey])
+	if cm.Data[EndpointDataKey] != "a=1.2.3.1:6790,b=1.2.3.2:6790" {
+		assert.Equal(t, "b=1.2.3.2:6790,a=1.2.3.1:6790", cm.Data[EndpointDataKey])
 	}
 
-	// Because mon2 isn't in the MonInQuorumResponse() this will create a mon11
+	// Because the mon a isn't in the MonInQuorumResponse() this will create a new mon
+	delete(c.mapping.Node, "rook-ceph-mon-a")
 	err = c.checkHealth()
 	assert.Nil(t, err)
 
 	// recheck that the "not found" mon has been replaced with a new one
 	cm, err = c.context.Clientset.CoreV1().ConfigMaps(c.Namespace).Get(EndpointConfigMapName, metav1.GetOptions{})
 	assert.Nil(t, err)
-	if cm.Data[EndpointDataKey] == "rook-ceph-mon1=1.2.3.1:6790,rook-ceph-mon11=:6790" {
-		assert.Equal(t, "rook-ceph-mon1=1.2.3.1:6790,rook-ceph-mon11=:6790", cm.Data[EndpointDataKey])
-	} else {
-		assert.Equal(t, "rook-ceph-mon11=:6790,rook-ceph-mon1=1.2.3.1:6790", cm.Data[EndpointDataKey])
+	if cm.Data[EndpointDataKey] != "b=1.2.3.2:6790,f=:6790" {
+		assert.Equal(t, "f=:6790,b=1.2.3.2:6790", cm.Data[EndpointDataKey])
 	}
 }
 
@@ -146,12 +141,12 @@ func TestCheckHealthTwoMonsOneNode(t *testing.T) {
 				resp := client.MonStatusResponse{Quorum: []int{0}}
 				resp.MonMap.Mons = []client.MonMapEntry{
 					{
-						Name:    "rook-ceph-mon1",
+						Name:    "a",
 						Rank:    0,
 						Address: "1.2.3.4",
 					},
 					{
-						Name:    "rook-ceph-mon3",
+						Name:    "c",
 						Rank:    0,
 						Address: "1.2.3.4",
 					},
@@ -177,24 +172,27 @@ func TestCheckHealthTwoMonsOneNode(t *testing.T) {
 	defer os.RemoveAll(c.context.ConfigDir)
 
 	// add two mons to the mapping on node0
-	c.mapping.Node["rook-ceph-mon1"] = &NodeInfo{
+	c.mapping.Node["a"] = &NodeInfo{
 		Name:     "node0",
 		Hostname: "mynode0",
 		Address:  "0.0.0.0",
 	}
-	c.mapping.Node["rook-ceph-mon2"] = &NodeInfo{
+	c.mapping.Node["b"] = &NodeInfo{
 		Name:     "node0",
 		Hostname: "mynode0",
 		Address:  "0.0.0.0",
 	}
-	c.maxMonID = 2
+	c.maxMonID = 1
 	c.saveMonConfig()
 
-	for i := 1; i <= 2; i++ {
-		rs := c.makeReplicaSet(&monConfig{Name: fmt.Sprintf("mon%d", i)}, "node0")
+	monNames := []string{"a", "b"}
+	for i := 0; i < len(monNames); i++ {
+		prefix := appName + "-"
+		name := monNames[i]
+		rs := c.makeReplicaSet(&monConfig{Name: prefix + name, DaemonName: name}, "node0")
 		_, err := clientset.ExtensionsV1beta1().ReplicaSets(c.Namespace).Create(rs)
 		assert.Nil(t, err)
-		po := c.makeMonPod(&monConfig{Name: fmt.Sprintf("mon%d", i)}, "node0")
+		po := c.makeMonPod(&monConfig{Name: prefix + name, DaemonName: name}, "node0")
 		_, err = clientset.CoreV1().Pods(c.Namespace).Create(po)
 		assert.Nil(t, err)
 	}
@@ -202,10 +200,10 @@ func TestCheckHealthTwoMonsOneNode(t *testing.T) {
 	// initial health check should already see that there is more than one mon on one node (node1)
 	_, err := c.checkMonsOnSameNode()
 	assert.Nil(t, err)
-	assert.Equal(t, "node0", c.mapping.Node["rook-ceph-mon1"].Name)
-	assert.Equal(t, "node0", c.mapping.Node["rook-ceph-mon2"].Name)
-	assert.Equal(t, "mynode0", c.mapping.Node["rook-ceph-mon1"].Hostname)
-	assert.Equal(t, "mynode0", c.mapping.Node["rook-ceph-mon2"].Hostname)
+	assert.Equal(t, "node0", c.mapping.Node["a"].Name)
+	assert.Equal(t, "node0", c.mapping.Node["b"].Name)
+	assert.Equal(t, "mynode0", c.mapping.Node["a"].Hostname)
+	assert.Equal(t, "mynode0", c.mapping.Node["b"].Hostname)
 
 	// add new node and check if the second mon gets failovered to it
 	n := &v1.Node{
@@ -229,9 +227,10 @@ func TestCheckHealthTwoMonsOneNode(t *testing.T) {
 	_, err = c.checkMonsOnSameNode()
 	assert.Nil(t, err)
 
-	// check that mon rook-ceph-mon3 exists
-	assert.NotNil(t, c.mapping.Node["rook-ceph-mon3"])
-	assert.Equal(t, "node2", c.mapping.Node["rook-ceph-mon3"].Name)
+	// check that mon rook-ceph-mon-c exists
+	logger.Infof("mapping: %+v", c.mapping.Node)
+	assert.NotNil(t, c.mapping.Node["rook-ceph-mon-c"])
+	assert.Equal(t, "node2", c.mapping.Node["rook-ceph-mon-c"].Name)
 
 	// check if mon2 has been deleted
 	var rsses *v1beta1.ReplicaSetList
@@ -240,13 +239,13 @@ func TestCheckHealthTwoMonsOneNode(t *testing.T) {
 
 	deleted := false
 	for _, rs := range rsses.Items {
-		if rs.Name == "rook-ceph-mon1" || rs.Name == "rook-ceph-mon3" {
+		if rs.Name == "rook-ceph-mon-a" || rs.Name == "rook-ceph-mon-c" {
 			deleted = true
 		} else {
 			deleted = false
 		}
 	}
-	assert.Equal(t, true, deleted, "rook-ceph-mon2 not failovered/deleted after health check")
+	assert.Equal(t, true, deleted, "rook-ceph-mon-b not failovered/deleted after health check")
 
 	// enable different ceph mon map output
 	executorNextMons = true
@@ -260,7 +259,7 @@ func TestCheckHealthTwoMonsOneNode(t *testing.T) {
 	for _, rs := range rsses.Items {
 		// both mons should always be on the same node as in this test due to the order
 		//the mons are processed in the loop
-		if (rs.Name == "rook-ceph-mon1" && rs.Spec.Template.Spec.NodeSelector[apis.LabelHostname] == "node1") || (rs.Name != "rook-ceph-mon3" && rs.Spec.Template.Spec.NodeSelector[apis.LabelHostname] == "node2") {
+		if (rs.Name == "rook-ceph-mon-a" && rs.Spec.Template.Spec.NodeSelector[apis.LabelHostname] == "node1") || (rs.Name != "rook-ceph-mon-c" && rs.Spec.Template.Spec.NodeSelector[apis.LabelHostname] == "node2") {
 			assert.Fail(t, fmt.Sprintf("mon %s shouldn't exist", rs.Name))
 		}
 	}
@@ -287,14 +286,15 @@ func TestCheckMonsValid(t *testing.T) {
 	defer os.RemoveAll(c.context.ConfigDir)
 
 	// add two mons to the mapping on node0
-	c.mapping.Node["rook-ceph-mon1"] = &NodeInfo{
+	c.mapping.Node["rook-ceph-mon-a"] = &NodeInfo{
 		Name:    "node0",
 		Address: "0.0.0.0",
 	}
-	c.mapping.Node["rook-ceph-mon2"] = &NodeInfo{
+	c.mapping.Node["rook-ceph-mon-b"] = &NodeInfo{
 		Name:    "node1",
 		Address: "0.0.0.0",
 	}
+	c.maxMonID = 1
 
 	// add three nodes
 	for i := 0; i < 3; i++ {
@@ -319,8 +319,8 @@ func TestCheckMonsValid(t *testing.T) {
 
 	_, err := c.checkMonsOnValidNodes()
 	assert.Nil(t, err)
-	assert.Equal(t, "node0", c.mapping.Node["rook-ceph-mon1"].Name)
-	assert.Equal(t, "node1", c.mapping.Node["rook-ceph-mon2"].Name)
+	assert.Equal(t, "node0", c.mapping.Node["rook-ceph-mon-a"].Name)
+	assert.Equal(t, "node1", c.mapping.Node["rook-ceph-mon-b"].Name)
 
 	// set node1 unschedulable and check that mon2 gets failovered to be mon3 to node2
 	node0, err := c.context.Clientset.CoreV1().Nodes().Get("node0", metav1.GetOptions{})
@@ -331,7 +331,8 @@ func TestCheckMonsValid(t *testing.T) {
 
 	// add the pods so the getNodesInUse() works correctly
 	for i := 1; i <= 2; i++ {
-		po := c.makeMonPod(&monConfig{Name: fmt.Sprintf("mon%d", i)}, fmt.Sprintf("node%d", i-1))
+		name := fmt.Sprintf("mon%d", i)
+		po := c.makeMonPod(&monConfig{Name: name, DaemonName: name}, fmt.Sprintf("node%d", i-1))
 		_, err = clientset.CoreV1().Pods(c.Namespace).Create(po)
 		assert.Nil(t, err)
 	}
@@ -340,11 +341,12 @@ func TestCheckMonsValid(t *testing.T) {
 	assert.Nil(t, err)
 
 	assert.Len(t, c.mapping.Node, 2)
-	assert.Nil(t, c.mapping.Node["rook-ceph-mon1"])
+	logger.Infof("mapping: %+v", c.mapping)
+	assert.Nil(t, c.mapping.Node["rook-ceph-mon-a"])
 	// the new mon should always be on the empty node2
-	// the failovered mon's name is "rook-ceph-mon0"
-	assert.Equal(t, "node2", c.mapping.Node["rook-ceph-mon0"].Name)
-	assert.Equal(t, "node1", c.mapping.Node["rook-ceph-mon2"].Name)
+	// the failovered mon's name is "rook-ceph-mon-a"
+	assert.Equal(t, "node2", c.mapping.Node["rook-ceph-mon-c"].Name)
+	assert.Equal(t, "node1", c.mapping.Node["rook-ceph-mon-b"].Name)
 }
 
 func TestCheckLessMonsStartNewMons(t *testing.T) {
@@ -363,12 +365,12 @@ func TestCheckLessMonsStartNewMons(t *testing.T) {
 	}
 	c := New(context, "ns", "", "myversion", cephv1beta1.MonSpec{Count: 5, AllowMultiplePerNode: true},
 		rookalpha.Placement{}, false, v1.ResourceRequirements{}, metav1.OwnerReference{})
-	c.maxMonID = 1
-	c.clusterInfo = test.CreateConfigDir(1)
+	c.maxMonID = 0
+	c.clusterInfo = test.CreateConfigDir(0)
 	c.waitForStart = false
 	defer os.RemoveAll(c.context.ConfigDir)
 
 	err := c.checkHealth()
 	assert.Nil(t, err)
-	assert.Equal(t, 5, len(c.clusterInfo.Monitors))
+	assert.Equal(t, 5, len(c.clusterInfo.Monitors), fmt.Sprintf("mons: %v", c.clusterInfo.Monitors))
 }

pkg/operator/ceph/cluster/mon/mon.go

@@ -21,6 +21,7 @@ import (
 	"encoding/json"
 	"fmt"
 	"strconv"
+	"strings"
 	"time"
 
 	"github.com/coreos/pkg/capnslog"
@@ -71,7 +72,6 @@ type Cluster struct {
 	Namespace            string
 	Keyring              string
 	Version              string
-	MasterHost           string
 	Size                 int
 	AllowMultiplePerNode bool
 	Port                 int32
@@ -91,9 +91,10 @@ type Cluster struct {
 
 // monConfig for a single monitor
 type monConfig struct {
-	Name     string
-	PublicIP string
-	Port     int32
+	Name       string
+	DaemonName string
+	PublicIP   string
+	Port       int32
 }
 
 // Mapping mon node and port mapping
@@ -216,18 +217,33 @@ func (c *Cluster) initMonConfig(size int) []*monConfig {
 
 	// initialize the mon pod info for mons that have been previously created
 	for _, monitor := range c.clusterInfo.Monitors {
-		mons = append(mons, &monConfig{Name: monitor.Name, Port: int32(mon.DefaultPort)})
+		mons = append(mons, &monConfig{Name: monitor.Name, DaemonName: daemonName(monitor.Name), Port: int32(mon.DefaultPort)})
 	}
 
 	// initialize mon info if we don't have enough mons (at first startup)
 	for i := len(c.clusterInfo.Monitors); i < size; i++ {
 		c.maxMonID++
-		mons = append(mons, &monConfig{Name: fmt.Sprintf("%s%d", appName, c.maxMonID), Port: int32(mon.DefaultPort)})
+		mons = append(mons, newMonConfig(c.maxMonID))
 	}
 
 	return mons
 }
 
+func newMonConfig(monID int) *monConfig {
+	daemonName := indexToName(monID)
+	return &monConfig{Name: fmt.Sprintf("%s-%s", appName, daemonName), DaemonName: daemonName, Port: int32(mon.DefaultPort)}
+}
+
+// Extract the daemon name from the full deployment name.
+func daemonName(fullName string) string {
+	prefix := appName + "-"
+	if strings.HasPrefix(fullName, prefix) {
+		return fullName[len(prefix):]
+	}
+	// If the deployment name did not have the prefix, we have a legacy daemon name that was in the form rook-ceph-mon0 and should not change
+	return fullName
+}
+
 func (c *Cluster) initMonIPs(mons []*monConfig) error {
 	for _, m := range mons {
 		if c.HostNetwork {
@@ -244,14 +260,14 @@ func (c *Cluster) initMonIPs(mons []*monConfig) error {
 			}
 			m.PublicIP = serviceIP
 		}
-		c.clusterInfo.Monitors[m.Name] = mon.ToCephMon(m.Name, m.PublicIP, m.Port)
+		c.clusterInfo.Monitors[m.DaemonName] = mon.ToCephMon(m.DaemonName, m.PublicIP, m.Port)
 	}
 
 	return nil
 }
 
 func (c *Cluster) createService(mon *monConfig) (string, error) {
-	labels := c.getLabels(mon.Name)
+	labels := c.getLabels(mon.DaemonName)
 	s := &v1.Service{
 		ObjectMeta: metav1.ObjectMeta{
 			Name:   mon.Name,
@@ -290,7 +306,7 @@ func (c *Cluster) createService(mon *monConfig) (string, error) {
 		return "", nil
 	}
 
-	logger.Infof("mon %s running at %s:%d", mon.Name, s.Spec.ClusterIP, mon.Port)
+	logger.Infof("mon %s running at %s:%d", mon.DaemonName, s.Spec.ClusterIP, mon.Port)
 	return s.Spec.ClusterIP, nil
 }
 
@@ -378,7 +394,7 @@ func (c *Cluster) waitForMonsToJoin(mons []*monConfig) error {
 
 	starting := []string{}
 	for _, m := range mons {
-		starting = append(starting, m.Name)
+		starting = append(starting, m.DaemonName)
 	}
 
 	// wait for the monitors to join quorum
@@ -535,7 +551,7 @@ func (c *Cluster) startMon(m *monConfig, hostname string) error {
 }
 
 func waitForQuorumWithMons(context *clusterd.Context, clusterName string, mons []string) error {
-	logger.Infof("waiting for mon quorum")
+	logger.Infof("waiting for mon quorum with %v", mons)
 
 	// wait for monitors to establish quorum
 	retryCount := 0

pkg/operator/ceph/cluster/mon/mon_test.go

@@ -84,6 +84,14 @@ func newCluster(context *clusterd.Context, namespace string, hostNetwork bool, r
 	}
 }
 
+func TestExtractDaemonName(t *testing.T) {
+	assert.Equal(t, "a", daemonName("rook-ceph-mon-a"))
+	assert.Equal(t, "b", daemonName("rook-ceph-mon-b"))
+	assert.Equal(t, "zz", daemonName("rook-ceph-mon-zz"))
+	assert.Equal(t, "rook-ceph-mon0", daemonName("rook-ceph-mon0"))
+	assert.Equal(t, "rook-ceph-mon123", daemonName("rook-ceph-mon123"))
+}
+
 func TestStartMonPods(t *testing.T) {
 	namespace := "ns"
 	context := newTestStartCluster(namespace)
@@ -151,7 +159,7 @@ func validateStart(t *testing.T, c *Cluster) {
 	assert.Equal(t, 4, len(s.StringData))
 
 	// there is only one pod created. the other two won't be created since the first one doesn't start
-	_, err = c.context.Clientset.Extensions().ReplicaSets(c.Namespace).Get("rook-ceph-mon0", metav1.GetOptions{})
+	_, err = c.context.Clientset.Extensions().ReplicaSets(c.Namespace).Get("rook-ceph-mon-a", metav1.GetOptions{})
 	assert.Nil(t, err)
 }
 
@@ -170,14 +178,14 @@ func TestSaveMonEndpoints(t *testing.T) {
 
 	cm, err := c.context.Clientset.CoreV1().ConfigMaps(c.Namespace).Get(EndpointConfigMapName, metav1.GetOptions{})
 	assert.Nil(t, err)
-	assert.Equal(t, "rook-ceph-mon1=1.2.3.1:6790", cm.Data[EndpointDataKey])
+	assert.Equal(t, "a=1.2.3.1:6790", cm.Data[EndpointDataKey])
 	assert.Equal(t, `{"node":{},"port":{}}`, cm.Data[MappingKey])
 	assert.Equal(t, "-1", cm.Data[MaxMonIDKey])
 
 	// update the config map
-	c.clusterInfo.Monitors["rook-ceph-mon1"].Endpoint = "2.3.4.5:6790"
+	c.clusterInfo.Monitors["a"].Endpoint = "2.3.4.5:6790"
 	c.maxMonID = 2
-	c.mapping.Node["rook-ceph-mon1"] = &NodeInfo{
+	c.mapping.Node["a"] = &NodeInfo{
 		Name:     "node0",
 		Address:  "1.1.1.1",
 		Hostname: "myhost",
@@ -188,8 +196,8 @@ func TestSaveMonEndpoints(t *testing.T) {
 
 	cm, err = c.context.Clientset.CoreV1().ConfigMaps(c.Namespace).Get(EndpointConfigMapName, metav1.GetOptions{})
 	assert.Nil(t, err)
-	assert.Equal(t, "rook-ceph-mon1=2.3.4.5:6790", cm.Data[EndpointDataKey])
-	assert.Equal(t, `{"node":{"rook-ceph-mon1":{"Name":"node0","Hostname":"myhost","Address":"1.1.1.1"}},"port":{"node0":12345}}`, cm.Data[MappingKey])
+	assert.Equal(t, "a=2.3.4.5:6790", cm.Data[EndpointDataKey])
+	assert.Equal(t, `{"node":{"a":{"Name":"node0","Hostname":"myhost","Address":"1.1.1.1"}},"port":{"node0":12345}}`, cm.Data[MappingKey])
 	assert.Equal(t, "2", cm.Data[MaxMonIDKey])
 }
 
@@ -210,23 +218,23 @@ func TestMonInQuourm(t *testing.T) {
 	assert.False(t, monInQuorum(entry, quorum))
 }
 
-func TestMonID(t *testing.T) {
+func TestNameToIndex(t *testing.T) {
 	// invalid
-	id, err := getMonID("m")
+	id, err := fullNameToIndex("m")
 	assert.NotNil(t, err)
 	assert.Equal(t, -1, id)
-	id, err = getMonID("mon")
+	id, err = fullNameToIndex("mon")
 	assert.NotNil(t, err)
 	assert.Equal(t, -1, id)
-	id, err = getMonID("rook-ceph-monitor0")
+	id, err = fullNameToIndex("rook-ceph-monitor0")
 	assert.NotNil(t, err)
 	assert.Equal(t, -1, id)
 
 	// valid
-	id, err = getMonID("rook-ceph-mon0")
+	id, err = fullNameToIndex("rook-ceph-mon-a")
 	assert.Nil(t, err)
 	assert.Equal(t, 0, id)
-	id, err = getMonID("rook-ceph-mon123")
+	id, err = fullNameToIndex("rook-ceph-mon123")
 	assert.Nil(t, err)
 	assert.Equal(t, 123, id)
 }
@@ -447,14 +455,17 @@ func TestHostNetworkPortIncrease(t *testing.T) {
 
 	mons := []*monConfig{
 		{
-			Name: "rook-ceph-mon1",
-			Port: cephmon.DefaultPort,
+			Name:       "rook-ceph-mon-a",
+			DaemonName: "a",
+			Port:       cephmon.DefaultPort,
 		},
 		{
-			Name: "rook-ceph-mon2",
-			Port: cephmon.DefaultPort,
+			Name:       "rook-ceph-mon-b",
+			DaemonName: "b",
+			Port:       cephmon.DefaultPort,
 		},
 	}
+	c.maxMonID = 1
 
 	err = c.assignMons(mons)
 	assert.Nil(t, err)
@@ -462,11 +473,11 @@ func TestHostNetworkPortIncrease(t *testing.T) {
 	err = c.initMonIPs(mons)
 	assert.Nil(t, err)
 
-	assert.Equal(t, node.Name, c.mapping.Node["rook-ceph-mon1"].Name)
-	assert.Equal(t, node.Name, c.mapping.Node["rook-ceph-mon2"].Name)
+	assert.Equal(t, node.Name, c.mapping.Node["rook-ceph-mon-a"].Name)
+	assert.Equal(t, node.Name, c.mapping.Node["rook-ceph-mon-b"].Name)
 
-	sEndpoint := strings.Split(c.clusterInfo.Monitors["rook-ceph-mon1"].Endpoint, ":")
+	sEndpoint := strings.Split(c.clusterInfo.Monitors["a"].Endpoint, ":")
 	assert.Equal(t, strconv.Itoa(cephmon.DefaultPort), sEndpoint[1])
-	sEndpoint = strings.Split(c.clusterInfo.Monitors["rook-ceph-mon2"].Endpoint, ":")
+	sEndpoint = strings.Split(c.clusterInfo.Monitors["b"].Endpoint, ":")
 	assert.Equal(t, strconv.Itoa(cephmon.DefaultPort+1), sEndpoint[1])
 }

pkg/operator/ceph/cluster/mon/spec.go

@@ -69,6 +69,7 @@ func (c *Cluster) makeReplicaSet(config *monConfig, hostname string) *extensions
 		ObjectMeta: metav1.ObjectMeta{
 			Name:      config.Name,
 			Namespace: c.Namespace,
+			Labels:    c.getLabels(config.DaemonName),
 		},
 	}
 	k8sutil.SetOwnerRef(c.context.Clientset, c.Namespace, &rs.ObjectMeta, &c.ownerRef)
@@ -115,7 +116,7 @@ func (c *Cluster) makeMonPod(config *monConfig, hostname string) *v1.Pod {
 		ObjectMeta: metav1.ObjectMeta{
 			Name:        config.Name,
 			Namespace:   c.Namespace,
-			Labels:      c.getLabels(config.Name),
+			Labels:      c.getLabels(config.DaemonName),
 			Annotations: map[string]string{},
 		},
 		Spec: podSpec,
@@ -136,7 +137,7 @@ func (c *Cluster) monContainer(config *monConfig, fsid string) v1.Container {
 			"ceph",
 			"mon",
 			fmt.Sprintf("--config-dir=%s", k8sutil.DataDir),
-			fmt.Sprintf("--name=%s", config.Name),
+			fmt.Sprintf("--name=%s", config.DaemonName),
 			fmt.Sprintf("--port=%d", config.Port),
 			fmt.Sprintf("--fsid=%s", fsid),
 		},

pkg/operator/ceph/cluster/mon/spec_test.go

@@ -48,11 +48,12 @@ func testPodSpec(t *testing.T, dataDir string) {
 			},
 		}, metav1.OwnerReference{})
 	c.clusterInfo = testop.CreateConfigDir(0)
-	config := &monConfig{Name: "rook-ceph-mon0", Port: 6790}
+	name := "rook-ceph-mon-a"
+	config := &monConfig{Name: name, DaemonName: name, Port: 6790}
 
 	pod := c.makeMonPod(config, "foo")
 	assert.NotNil(t, pod)
-	assert.Equal(t, "rook-ceph-mon0", pod.Name)
+	assert.Equal(t, "rook-ceph-mon-a", pod.Name)
 	assert.Equal(t, v1.RestartPolicyAlways, pod.Spec.RestartPolicy)
 	assert.Equal(t, 2, len(pod.Spec.Volumes))
 	assert.Equal(t, "rook-data", pod.Spec.Volumes[0].Name)
@@ -65,7 +66,7 @@ func testPodSpec(t *testing.T, dataDir string) {
 		assert.Equal(t, dataDir, pod.Spec.Volumes[0].HostPath.Path)
 	}
 
-	assert.Equal(t, "rook-ceph-mon0", pod.ObjectMeta.Name)
+	assert.Equal(t, "rook-ceph-mon-a", pod.ObjectMeta.Name)
 	assert.Equal(t, appName, pod.ObjectMeta.Labels["app"])
 	assert.Equal(t, c.Namespace, pod.ObjectMeta.Labels["mon_cluster"])
 
@@ -79,7 +80,7 @@ func testPodSpec(t *testing.T, dataDir string) {
 	assert.Equal(t, "ceph", cont.Args[0])
 	assert.Equal(t, "mon", cont.Args[1])
 	assert.Equal(t, "--config-dir=/var/lib/rook", cont.Args[2])
-	assert.Equal(t, "--name=rook-ceph-mon0", cont.Args[3])
+	assert.Equal(t, "--name="+name, cont.Args[3])
 	assert.Equal(t, "--port=6790", cont.Args[4])
 	assert.Equal(t, fmt.Sprintf("--fsid=%s", c.clusterInfo.FSID), cont.Args[5])
 

pkg/operator/ceph/cluster/mon/util.go

@@ -39,6 +39,10 @@ import (
 	"k8s.io/client-go/kubernetes"
 )
 
+const (
+	maxPerChar = 26
+)
+
 // LoadClusterInfo constructs or loads a clusterinfo and returns it along with the maxMonID
 func LoadClusterInfo(context *clusterd.Context, namespace string) (*mon.ClusterInfo, int, *Mapping, error) {
 	return CreateOrLoadClusterInfo(context, namespace, nil)
@@ -135,7 +139,7 @@ func loadMonConfig(clientset kubernetes.Interface, namespace string) (map[string
 
 	// Make sure the max id is consistent with the current monitors
 	for _, m := range monEndpointMap {
-		id, _ := getMonID(m.Name)
+		id, _ := fullNameToIndex(m.Name)
 		if maxMonID < id {
 			maxMonID = id
 		}
@@ -150,18 +154,6 @@ func loadMonConfig(clientset kubernetes.Interface, namespace string) (map[string
 	return monEndpointMap, maxMonID, monMapping, nil
 }
 
-// get the ID of a monitor from its name
-func getMonID(name string) (int, error) {
-	if strings.Index(name, appName) != 0 || len(name) < len(appName) {
-		return -1, fmt.Errorf("unexpected mon name")
-	}
-	id, err := strconv.Atoi(name[len(appName):])
-	if err != nil {
-		return -1, err
-	}
-	return id, nil
-}
-
 func createClusterAccessSecret(clientset kubernetes.Interface, namespace string, clusterInfo *mon.ClusterInfo, ownerRef *metav1.OwnerReference) error {
 	logger.Infof("creating mon secrets for a new cluster")
 	var err error
@@ -265,3 +257,56 @@ func extractKey(contents string) (string, error) {
 	}
 	return secret, nil
 }
+
+// convert an index to a mon name based on as few letters of the alphabet as possible
+func indexToName(index int) string {
+	var result string
+	for {
+		i := index % maxPerChar
+		c := 'z' - maxPerChar + i + 1
+		result = fmt.Sprintf("%c%s", c, result)
+		if index < maxPerChar {
+			break
+		}
+		// subtract 1 since the character conversion is zero-based
+		index = (index / maxPerChar) - 1
+	}
+	return result
+}
+
+// convert the mon name to the numeric mon ID
+func fullNameToIndex(name string) (int, error) {
+	prefix := appName + "-"
+	if strings.Index(name, prefix) != -1 && len(prefix) < len(name) {
+		return nameToIndex(name[len(prefix)+1:])
+	}
+
+	// attempt to parse the legacy mon name
+	legacyPrefix := appName
+	if strings.Index(name, legacyPrefix) == -1 || len(name) < len(appName) {
+		return -1, fmt.Errorf("unexpected mon name")
+	}
+	id, err := strconv.Atoi(name[len(legacyPrefix):])
+	if err != nil {
+		return -1, err
+	}
+	return id, nil
+}
+
+func nameToIndex(name string) (int, error) {
+	factor := 1
+	for i := 1; i < len(name); i++ {
+		factor *= maxPerChar
+	}
+	var result int
+	for _, c := range name {
+		charVal := int('z' - c + 1)
+		if charVal < 0 || charVal >= maxPerChar {
+			return -1, fmt.Errorf("invalid char '%c' in %s", c, name)
+		}
+		result += charVal * factor
+		factor /= maxPerChar
+	}
+	return result, nil
+
+}

pkg/operator/ceph/cluster/mon/util_test.go

+/*
+Copyright 2018 The Rook Authors. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+	http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+package mon
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestConvertMonID(t *testing.T) {
+	assert.Equal(t, "a", indexToName(0))
+	assert.Equal(t, "b", indexToName(1))
+	assert.Equal(t, "c", indexToName(2))
+	assert.Equal(t, "z", indexToName(25))
+	assert.Equal(t, "aa", indexToName(26))
+	assert.Equal(t, "ab", indexToName(27))
+	assert.Equal(t, "ac", indexToName(28))
+	assert.Equal(t, "az", indexToName(51))
+	assert.Equal(t, "ba", indexToName(52))
+	assert.Equal(t, "bb", indexToName(53))
+	assert.Equal(t, "bz", indexToName(77))
+	assert.Equal(t, "ca", indexToName(78))
+	assert.Equal(t, "za", indexToName(676))
+	assert.Equal(t, "aaa", indexToName(702))
+	assert.Equal(t, "aaz", indexToName(727))
+	assert.Equal(t, "aba", indexToName(728))
+}

pkg/operator/test/info.go

@@ -24,7 +24,7 @@ import (
 )
 
 // CreateConfigDir creates a test cluster
-func CreateConfigDir(mons int) *mon.ClusterInfo {
+func CreateConfigDir(monCount int) *mon.ClusterInfo {
 	c := &mon.ClusterInfo{
 		FSID:          "12345",
 		Name:          "default",
@@ -32,11 +32,12 @@ func CreateConfigDir(mons int) *mon.ClusterInfo {
 		AdminSecret:   "adminsecret",
 		Monitors:      map[string]*mon.CephMonitorConfig{},
 	}
-	for i := 1; i <= mons; i++ {
-		id := fmt.Sprintf("rook-ceph-mon%d", i)
+	mons := []string{"a", "b", "c", "d", "e"}
+	for i := 0; i < monCount; i++ {
+		id := mons[i]
 		c.Monitors[id] = &mon.CephMonitorConfig{
 			Name:     id,
-			Endpoint: fmt.Sprintf("1.2.3.%d:6790", i),
+			Endpoint: fmt.Sprintf("1.2.3.%d:6790", i+1),
 		}
 	}
 	return c