ceph: improve upgrade procedure
When a cluster is updated with a different image version, this triggers a serialized restart of all the pods. Prior to this commit, no safety check were performed and rook was hoping for the best outcome. Now before doing restarting a daemon we check it can be restarted. Once it's restarted we also check we can pursue with the rest of the platform. For instance, with monitors we check that they are in quorum, for OSD we check that PGs are clean and for MDS we make sure they are all active. Fixes: https://github.com/rook/rook/issues/2889 Signed-off-by: Sébastien Han <seb@redhat.com>
This commit is contained in:
1 parent
487ded50f3
commit
9824bd2ff9
14 files changed
+580
-52
No files matched your search
@@ -52,6 +52,7 @@ type CephStatus struct {
|
||||
} `json:"osdmap"`
|
||||
PgMap PgMap `json:"pgmap"`
|
||||
MgrMap MgrMap `json:"mgrmap"`
|
||||
Fsmap Fsmap `json:"fsmap"`
|
||||
}
|
||||
|
||||
type HealthStatus struct {
|
||||
@@ -123,6 +124,23 @@ type PgStateEntry struct {
|
||||
Count int `json:"count"`
|
||||
}
|
||||
|
||||
// Fsmap is a struct representing the filesystem map
|
||||
type Fsmap struct {
|
||||
Epoch int `json:"epoch"`
|
||||
ID int `json:"id"`
|
||||
Up int `json:"up"`
|
||||
In int `json:"in"`
|
||||
Max int `json:"max"`
|
||||
ByRank []struct {
|
||||
FilesystemID int `json:"filesystem_id"`
|
||||
Rank int `json:"rank"`
|
||||
Name string `json:"name"`
|
||||
Status string `json:"status"`
|
||||
Gid int `json:"gid"`
|
||||
} `json:"by_rank"`
|
||||
UpStandby int `json:"up:standby"`
|
||||
}
|
||||
|
||||
func Status(context *clusterd.Context, clusterName string, debug bool) (CephStatus, error) {
|
||||
args := []string{"status"}
|
||||
cmd := NewCephCommand(context, clusterName, args)
|
||||
@@ -171,3 +189,50 @@ func isClusterClean(status CephStatus) error {
|
||||
|
||||
return fmt.Errorf("cluster is not fully clean. PGs: %+v", status.PgMap.PgsByState)
|
||||
}
|
||||
|
||||
// getMDSRank returns the rank of a given MDS
|
||||
func getMDSRank(status CephStatus, clusterName, fsName string) (int, error) {
|
||||
// dummy rank
|
||||
mdsRank := -1000
|
||||
for r := range status.Fsmap.ByRank {
|
||||
if status.Fsmap.ByRank[r].Name == fsName {
|
||||
mdsRank = r
|
||||
}
|
||||
}
|
||||
// if the mds is not shown in the map one reason might be because it's in standby
|
||||
// if not in standby there is something else going wron
|
||||
if mdsRank < 0 && status.Fsmap.UpStandby < 1 {
|
||||
// it might seem strange to log an error since this could be a warning too
|
||||
// it is a warning until we reach the timeout, this should give enough time to the mds to transtion its state
|
||||
// after the timeout we consider that the mds might be gone or the timeout was not long enough...
|
||||
return mdsRank, fmt.Errorf("mds %s not found in fsmap, this likely means mdss are transitioning between active and standby states", fsName)
|
||||
}
|
||||
|
||||
return mdsRank, nil
|
||||
}
|
||||
|
||||
// MdsActiveOrStandbyReplay returns wether a given MDS is active or in standby
|
||||
func MdsActiveOrStandbyReplay(context *clusterd.Context, clusterName, fsName string) error {
|
||||
status, err := Status(context, clusterName, false)
|
||||
if err != nil {
|
||||
return fmt.Errorf("failed to get ceph status. %+v", err)
|
||||
}
|
||||
|
||||
mdsRank, err := getMDSRank(status, clusterName, fsName)
|
||||
if err != nil {
|
||||
return fmt.Errorf("%+v", err)
|
||||
}
|
||||
|
||||
// this MDS is in standby so let's return immediatly
|
||||
if mdsRank < 0 {
|
||||
logger.Infof("mds %s is in standby, nothing to check", fsName)
|
||||
return nil
|
||||
}
|
||||
|
||||
if status.Fsmap.ByRank[mdsRank].Status == "up:active" || status.Fsmap.ByRank[mdsRank].Status == "up:standby-replay" || status.Fsmap.ByRank[mdsRank].Status == "up:standby" {
|
||||
logger.Infof("mds %s is %s", fsName, status.Fsmap.ByRank[mdsRank].Status)
|
||||
return nil
|
||||
}
|
||||
|
||||
return fmt.Errorf("mds %s is %s, bad state", fsName, status.Fsmap.ByRank[mdsRank].Status)
|
||||
}
|
||||
@@ -27,6 +27,30 @@ const (
|
||||
CephStatusResponseRaw = `{"fsid":"613975f3-3025-4802-9de1-a2280b950e75","health":{"checks":{"OSD_DOWN":{"severity":"HEALTH_WARN","summary":{"message":"1 osds down"}},"OSD_HOST_DOWN":{"severity":"HEALTH_WARN","summary":{"message":"1 host (1 osds) down"}},"PG_AVAILABILITY":{"severity":"HEALTH_WARN","summary":{"message":"Reduced data availability: 101 pgs stale"}},"POOL_APP_NOT_ENABLED":{"severity":"HEALTH_WARN","summary":{"message":"application not enabled on 1 pool(s)"}}},"status":"HEALTH_WARN","overall_status":"HEALTH_WARN"},"election_epoch":12,"quorum":[0,1,2],"quorum_names":["rook-ceph-mon0","rook-ceph-mon2","rook-ceph-mon1"],"monmap":{"epoch":3,"fsid":"613975f3-3025-4802-9de1-a2280b950e75","modified":"2017-08-11 20:13:02.075679","created":"2017-08-11 20:12:35.314510","features":{"persistent":["kraken","luminous"],"optional":[]},"mons":[{"rank":0,"name":"rook-ceph-mon0","addr":"10.3.0.45:6789/0","public_addr":"10.3.0.45:6789/0"},{"rank":1,"name":"rook-ceph-mon2","addr":"10.3.0.249:6789/0","public_addr":"10.3.0.249:6789/0"},{"rank":2,"name":"rook-ceph-mon1","addr":"10.3.0.252:6789/0","public_addr":"10.3.0.252:6789/0"}]},"osdmap":{"osdmap":{"epoch":17,"num_osds":2,"num_up_osds":1,"num_in_osds":2,"full":false,"nearfull":true,"num_remapped_pgs":0}},"pgmap":{"pgs_by_state":[{"state_name":"stale+active+clean","count":101},{"state_name":"active+clean","count":99}],"num_pgs":200,"num_pools":2,"num_objects":243,"data_bytes":976793635,"bytes_used":13611479040,"bytes_avail":19825307648,"bytes_total":33436786688},"fsmap":{"epoch":1,"by_rank":[]},"mgrmap":{"epoch":3,"active_gid":14111,"active_name":"rook-ceph-mgr0","active_addr":"10.2.73.6:6800/9","available":true,"standbys":[],"modules":["restful","status"],"available_modules":["dashboard","prometheus","restful","status","zabbix"]},"servicemap":{"epoch":1,"modified":"0.000000","services":{}}}`
|
||||
)
|
||||
|
||||
var (
|
||||
// this JSON was generated from `ceph status -f json`, using Ceph Nautilus 14.2.1
|
||||
// It was chopped to only show what the tests are looking for
|
||||
statusFakeRaw = []byte(`{
|
||||
"fsmap": {
|
||||
"epoch": 13,
|
||||
"id": 1,
|
||||
"up": 1,
|
||||
"in": 1,
|
||||
"max": 1,
|
||||
"by_rank": [
|
||||
{
|
||||
"filesystem_id": 1,
|
||||
"rank": 0,
|
||||
"name": "myfs-b",
|
||||
"status": "up:active",
|
||||
"gid": 14716
|
||||
}
|
||||
],
|
||||
"up:standby": 1
|
||||
}
|
||||
}`)
|
||||
)
|
||||
|
||||
func TestStatusMarshal(t *testing.T) {
|
||||
var status CephStatus
|
||||
err := json.Unmarshal([]byte(CephStatusResponseRaw), &status)
|
||||
@@ -85,3 +109,12 @@ func TestIsClusterClean(t *testing.T) {
|
||||
err = isClusterClean(status)
|
||||
assert.NotNil(t, err)
|
||||
}
|
||||
|
||||
func TestGetMDSRank(t *testing.T) {
|
||||
var statusFake CephStatus
|
||||
json.Unmarshal(statusFakeRaw, &statusFake)
|
||||
|
||||
mdsRankFake, err := getMDSRank(statusFake, "rook-ceph", "myfs-b")
|
||||
assert.Nil(t, err)
|
||||
assert.Equal(t, 0, mdsRankFake)
|
||||
}
|
||||
@@ -18,35 +18,72 @@ package client
|
||||
|
||||
import (
|
||||
"encoding/json"
|
||||
"errors"
|
||||
"fmt"
|
||||
"strings"
|
||||
"time"
|
||||
|
||||
"github.com/rook/rook/pkg/clusterd"
|
||||
cephver "github.com/rook/rook/pkg/operator/ceph/version"
|
||||
"github.com/rook/rook/pkg/util"
|
||||
)
|
||||
|
||||
// CephDaemonsVersions is a structure that can be used to parsed the output of the 'ceph versions' command
|
||||
type CephDaemonsVersions struct {
|
||||
Mon map[string]int `json:"mon,omitempty"`
|
||||
Osd map[string]int `json:"osd,omitempty"`
|
||||
Mgr map[string]int `json:"mgr,omitempty"`
|
||||
Mds map[string]int `json:"mds,omitempty"`
|
||||
Overall map[string]int `json:"overall,omitempty"`
|
||||
Mon map[string]int `json:"mon,omitempty"`
|
||||
Mgr map[string]int `json:"mgr,omitempty"`
|
||||
Osd map[string]int `json:"osd,omitempty"`
|
||||
Rgw map[string]int `json:"rgw,omitempty"`
|
||||
Mds map[string]int `json:"mds,omitempty"`
|
||||
RbdMirror map[string]int `json:"rbd-mirror,omitempty"`
|
||||
Overall map[string]int `json:"overall,omitempty"`
|
||||
}
|
||||
|
||||
func getCephMonVersionString(context *clusterd.Context) (string, error) {
|
||||
output, err := context.Executor.ExecuteCommandWithOutput(false, "", "ceph", "version")
|
||||
var (
|
||||
// we don't perform any checks on these daemons
|
||||
// they don't have any "ok-to-stop" command implemented
|
||||
daemonNoCheck = []string{"mgr", "rgw", "rbd-mirror", "nfs"}
|
||||
)
|
||||
|
||||
func getCephMonVersionString(context *clusterd.Context, clusterName string) (string, error) {
|
||||
args := []string{"version"}
|
||||
command, args := FinalizeCephCommandArgs("ceph", args, context.ConfigDir, clusterName)
|
||||
|
||||
output, err := context.Executor.ExecuteCommandWithOutput(false, "", command, args...)
|
||||
if err != nil {
|
||||
return "", fmt.Errorf("failed to run ceph version: %+v", err)
|
||||
return "", fmt.Errorf("failed to run 'ceph version'. %+v", err)
|
||||
}
|
||||
logger.Debug(output)
|
||||
|
||||
return output, nil
|
||||
}
|
||||
|
||||
func getCephVersionsString(context *clusterd.Context) (string, error) {
|
||||
output, err := context.Executor.ExecuteCommandWithOutput(false, "", "ceph", "versions")
|
||||
func getAllCephDaemonVersionsString(context *clusterd.Context, clusterName string) (string, error) {
|
||||
args := []string{"versions"}
|
||||
command, args := FinalizeCephCommandArgs("ceph", args, context.ConfigDir, clusterName)
|
||||
|
||||
output, err := context.Executor.ExecuteCommandWithOutput(false, "", command, args...)
|
||||
if err != nil {
|
||||
return "", fmt.Errorf("failed to run ceph versions: %+v", err)
|
||||
return "", fmt.Errorf("failed to run 'ceph versions'. %+v", err)
|
||||
}
|
||||
logger.Debug(output)
|
||||
|
||||
return output, nil
|
||||
}
|
||||
|
||||
func getCephDaemonVersionString(context *clusterd.Context, deployment, clusterName string) (string, error) {
|
||||
daemonName, err := findDaemonName(deployment)
|
||||
if err != nil {
|
||||
return "", fmt.Errorf("%+v", err)
|
||||
}
|
||||
daemonID := findDaemonID(deployment)
|
||||
daemon := daemonName + "." + daemonID
|
||||
|
||||
args := []string{"tell", daemon, "version"}
|
||||
command, args := FinalizeCephCommandArgs("ceph", args, context.ConfigDir, clusterName)
|
||||
output, err := context.Executor.ExecuteCommandWithOutput(false, "", command, args...)
|
||||
if err != nil {
|
||||
return "", fmt.Errorf("failed to run ceph tell. %+v", err)
|
||||
}
|
||||
logger.Debug(output)
|
||||
|
||||
@@ -54,10 +91,10 @@ func getCephVersionsString(context *clusterd.Context) (string, error) {
|
||||
}
|
||||
|
||||
// GetCephMonVersion reports the Ceph version of all the monitors, or at least a majority with quorum
|
||||
func GetCephMonVersion(context *clusterd.Context) (*cephver.CephVersion, error) {
|
||||
output, err := getCephMonVersionString(context)
|
||||
func GetCephMonVersion(context *clusterd.Context, clusterName string) (*cephver.CephVersion, error) {
|
||||
output, err := getCephMonVersionString(context, clusterName)
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("failed to run ceph version: %+v", err)
|
||||
return nil, err
|
||||
}
|
||||
logger.Debug(output)
|
||||
|
||||
@@ -69,11 +106,27 @@ func GetCephMonVersion(context *clusterd.Context) (*cephver.CephVersion, error)
|
||||
return v, nil
|
||||
}
|
||||
|
||||
// GetCephVersions reports the Ceph version of each daemon in the cluster
|
||||
func GetCephVersions(context *clusterd.Context) (*CephDaemonsVersions, error) {
|
||||
output, err := getCephVersionsString(context)
|
||||
// GetCephDaemonVersion reports the Ceph version of a particular daemon
|
||||
func GetCephDaemonVersion(context *clusterd.Context, deployment, clusterName string) (*cephver.CephVersion, error) {
|
||||
output, err := getCephDaemonVersionString(context, deployment, clusterName)
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("failed to run ceph versions: %+v", err)
|
||||
return nil, fmt.Errorf("failed to run ceph tell. %+v", err)
|
||||
}
|
||||
logger.Debug(output)
|
||||
|
||||
v, err := cephver.ExtractCephVersion(output)
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("failed to extract ceph version. %+v", err)
|
||||
}
|
||||
|
||||
return v, nil
|
||||
}
|
||||
|
||||
// GetAllCephDaemonVersions reports the Ceph version of each daemon in the cluster
|
||||
func GetAllCephDaemonVersions(context *clusterd.Context, clusterName string) (*CephDaemonsVersions, error) {
|
||||
output, err := getAllCephDaemonVersionsString(context, clusterName)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
logger.Debug(output)
|
||||
|
||||
@@ -90,7 +143,7 @@ func GetCephVersions(context *clusterd.Context) (*CephDaemonsVersions, error) {
|
||||
func EnableMessenger2(context *clusterd.Context) error {
|
||||
_, err := context.Executor.ExecuteCommandWithOutput(false, "", "ceph", "mon", "enable-msgr2")
|
||||
if err != nil {
|
||||
return fmt.Errorf("failed to enable msgr2 protocol: %+v", err)
|
||||
return fmt.Errorf("failed to enable msgr2 protocol. %+v", err)
|
||||
}
|
||||
logger.Infof("successfully enabled msgr2 protocol")
|
||||
|
||||
@@ -104,6 +157,197 @@ func EnableNautilusOSD(context *clusterd.Context) error {
|
||||
return fmt.Errorf("failed to disallow pre-nautilus osds and enable all new nautilus-only functionality: %+v", err)
|
||||
}
|
||||
logger.Infof("successfully disallowed pre-nautilus osds and enabled all new nautilus-only functionality")
|
||||
return nil
|
||||
}
|
||||
|
||||
// OkToStop determines if it's ok to stop an upgrade
|
||||
func OkToStop(context *clusterd.Context, namespace, deployment, clusterName string, cephVersion cephver.CephVersion) error {
|
||||
daemonName, err := findDaemonName(deployment)
|
||||
if err != nil {
|
||||
logger.Warningf("%+v", err)
|
||||
return nil
|
||||
}
|
||||
|
||||
// The ok-to-stop command for mon and mds landed on 14.2.1
|
||||
// so we return nil if that Ceph version is not satisfied
|
||||
if !cephVersion.IsAtLeast(cephver.CephVersion{Major: 14, Minor: 2, Extra: 1}) {
|
||||
if daemonName != "osd" {
|
||||
return nil
|
||||
}
|
||||
}
|
||||
|
||||
versions, err := GetAllCephDaemonVersions(context, clusterName)
|
||||
if err != nil {
|
||||
return fmt.Errorf("failed to get ceph daemons versions. %+v", err)
|
||||
}
|
||||
|
||||
switch daemonName {
|
||||
// Trying to handle the case where a **single** mon is deployed and an upgrade is called
|
||||
case "mon":
|
||||
// if len(versions.Mon) > 1, this means we have different Ceph versions for some monitor(s).
|
||||
// This is fine, we can run the upgrade checks
|
||||
if len(versions.Mon) == 1 {
|
||||
// now trying to parse and find how many mons are presents
|
||||
// if we have less than 3 mons we skip the check and do best-effort
|
||||
// we do less than 3 because during the initial bootstrap the mon sequence is updated too
|
||||
// so running running the check on 2/3 mon fails
|
||||
// versions.Mon looks like this map[ceph version 15.0.0-12-g6c8fb92 (6c8fb920cb1d862f36ee852ed849a15f9a50bd68) octopus (dev):1]
|
||||
// now looping over a single element since we can't address the key directly (we don't know its name)
|
||||
for _, monCount := range versions.Mon {
|
||||
if monCount < 3 {
|
||||
logger.Infof("the cluster has less than 3 monitors, not performing upgrade check, running in best-effort")
|
||||
return nil
|
||||
}
|
||||
}
|
||||
}
|
||||
// Trying to handle the case where a **single** osd is deployed and an upgrade is called
|
||||
case "osd":
|
||||
// if len(versions.Osd) > 1, this means we have different Ceph versions osd(s)
|
||||
// This is fine, we can run the upgrade checks
|
||||
if len(versions.Osd) == 1 {
|
||||
// now trying to parse and find how many osds are presents
|
||||
// if we have less than 3 osds we skip the check and do best-effort
|
||||
for _, osdCount := range versions.Osd {
|
||||
if osdCount < 3 {
|
||||
logger.Infof("the cluster has less than 3 OSDs, not performing upgrade check, running in best-effort")
|
||||
return nil
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
// we don't implement any checks for mon, rgw and rbdmirror since:
|
||||
// - mon: the is done in the monitor code since it ensures all the mons are always in quorum before continuing
|
||||
// - rgw: the pod spec has a liveness probe so if the pod successfully start
|
||||
// - rbdmirror: you can chain as many as you want like mdss but there is no ok-to-stop logic yet
|
||||
err = okToStopDaemon(context, deployment, clusterName)
|
||||
if err != nil {
|
||||
return fmt.Errorf("failed to check if %s was ok to stop. %+v", deployment, err)
|
||||
}
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
// OkToContinue determines if it's ok to continue an upgrade
|
||||
func OkToContinue(context *clusterd.Context, namespace, deployment string) error {
|
||||
daemonName, err := findDaemonName(deployment)
|
||||
if err != nil {
|
||||
logger.Warningf("%+v", err)
|
||||
return nil
|
||||
}
|
||||
|
||||
// the mon case is handled directly in the deployment where the mon checks for quorum
|
||||
switch daemonName {
|
||||
case "osd":
|
||||
err := okToContinueOSDDaemon(context, namespace)
|
||||
if err != nil {
|
||||
return fmt.Errorf("failed to check if %s was ok to continue. %+v", deployment, err)
|
||||
}
|
||||
case "mds":
|
||||
err := okToContinueMDSDaemon(context, namespace, deployment)
|
||||
if err != nil {
|
||||
return fmt.Errorf("failed to check if %s was ok to continue. %+v", deployment, err)
|
||||
}
|
||||
}
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
func okToStopDaemon(context *clusterd.Context, deployment, clusterName string) error {
|
||||
daemonID := findDaemonID(deployment)
|
||||
daemonName, err := findDaemonName(deployment)
|
||||
if err != nil {
|
||||
logger.Warningf("%+v", err)
|
||||
return nil
|
||||
}
|
||||
|
||||
if !stringInSlice(daemonName, daemonNoCheck) {
|
||||
args := []string{daemonName, "ok-to-stop", daemonID}
|
||||
command, args := FinalizeCephCommandArgs("ceph", args, context.ConfigDir, clusterName)
|
||||
|
||||
output, err := context.Executor.ExecuteCommandWithOutput(false, "", command, args...)
|
||||
if err != nil {
|
||||
return fmt.Errorf("deployment %s cannot be stopped. %+v", deployment, err)
|
||||
}
|
||||
logger.Debugf("deployment %s is ok to be updated. %s", deployment, output)
|
||||
}
|
||||
|
||||
logger.Debugf("deployment %s is ok to be updated.", deployment)
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
// okToContinueOSDDaemon determines whether it's fine to go to the next osd during an upgrade
|
||||
// This basically makes sure all the PGs have settled
|
||||
func okToContinueOSDDaemon(context *clusterd.Context, namespace string) error {
|
||||
// Reconciliating PGs should not take too long so let's wait up to 10 minutes
|
||||
err := util.Retry(10, 60*time.Second, func() error {
|
||||
return IsClusterClean(context, namespace)
|
||||
})
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
// okToContinueMDSDaemon determines whether it's fine to go to the next mds during an upgrade
|
||||
// mostly a placeholder function for the future but since we have standby mds this shouldn't be needed
|
||||
func okToContinueMDSDaemon(context *clusterd.Context, namespace, deployment string) error {
|
||||
// wait for the MDS to be active again or in standby-replay
|
||||
err := util.Retry(10, 15*time.Second, func() error {
|
||||
return MdsActiveOrStandbyReplay(context, namespace, findFSName(deployment, namespace))
|
||||
})
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
func findFSName(deployment, namespace string) string {
|
||||
mdsID := findDaemonID(deployment)
|
||||
fsNameTrimSuffix := strings.TrimSuffix(deployment, "-"+mdsID)
|
||||
return strings.TrimPrefix(fsNameTrimSuffix, namespace+"-mds-")
|
||||
}
|
||||
|
||||
func findDaemonID(deployment string) string {
|
||||
daemonTrimPrefixSplit := strings.Split(deployment, "-")
|
||||
return daemonTrimPrefixSplit[len(daemonTrimPrefixSplit)-1]
|
||||
}
|
||||
|
||||
func findDaemonName(deployment string) (string, error) {
|
||||
err := errors.New("could not find daemon name, is this a new daemon?")
|
||||
|
||||
if strings.Contains(deployment, "mds") {
|
||||
return "mds", nil
|
||||
}
|
||||
if strings.Contains(deployment, "rgw") {
|
||||
return "rgw", nil
|
||||
}
|
||||
if strings.Contains(deployment, "mon") {
|
||||
return "mon", nil
|
||||
}
|
||||
if strings.Contains(deployment, "osd") {
|
||||
return "osd", nil
|
||||
}
|
||||
if strings.Contains(deployment, "mgr") {
|
||||
return "mgr", nil
|
||||
}
|
||||
if strings.Contains(deployment, "nfs") {
|
||||
return "nfs", nil
|
||||
}
|
||||
if strings.Contains(deployment, "rbd-mirror") {
|
||||
return "rbd-mirror", nil
|
||||
}
|
||||
|
||||
return "", fmt.Errorf("%+v from deployment %s", err, deployment)
|
||||
}
|
||||
|
||||
func stringInSlice(a string, list []string) bool {
|
||||
for _, b := range list {
|
||||
if b == a {
|
||||
return true
|
||||
}
|
||||
}
|
||||
return false
|
||||
}
|
||||
@@ -17,9 +17,11 @@ limitations under the License.
|
||||
package client
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
"testing"
|
||||
|
||||
"github.com/rook/rook/pkg/clusterd"
|
||||
cephver "github.com/rook/rook/pkg/operator/ceph/version"
|
||||
exectest "github.com/rook/rook/pkg/util/exec/test"
|
||||
"github.com/stretchr/testify/assert"
|
||||
)
|
||||
@@ -28,12 +30,11 @@ func TestGetCephMonVersionString(t *testing.T) {
|
||||
executor := &exectest.MockExecutor{}
|
||||
executor.MockExecuteCommandWithOutput = func(debug bool, name string, command string, args ...string) (string, error) {
|
||||
assert.Equal(t, "version", args[0])
|
||||
assert.Equal(t, 1, len(args))
|
||||
return "", nil
|
||||
}
|
||||
context := &clusterd.Context{Executor: executor}
|
||||
|
||||
_, err := getCephMonVersionString(context)
|
||||
_, err := getCephMonVersionString(context, "rook-ceph")
|
||||
assert.Nil(t, err)
|
||||
}
|
||||
|
||||
@@ -41,12 +42,26 @@ func TestGetCephMonVersionsString(t *testing.T) {
|
||||
executor := &exectest.MockExecutor{}
|
||||
executor.MockExecuteCommandWithOutput = func(debug bool, name string, command string, args ...string) (string, error) {
|
||||
assert.Equal(t, "versions", args[0])
|
||||
assert.Equal(t, 1, len(args))
|
||||
return "", nil
|
||||
}
|
||||
context := &clusterd.Context{Executor: executor}
|
||||
|
||||
_, err := getCephVersionsString(context)
|
||||
_, err := getAllCephDaemonVersionsString(context, "rook-ceph")
|
||||
assert.Nil(t, err)
|
||||
}
|
||||
|
||||
func TestGetCephDaemonVersionString(t *testing.T) {
|
||||
executor := &exectest.MockExecutor{}
|
||||
deployment := "rook-ceph-mds-a"
|
||||
executor.MockExecuteCommandWithOutput = func(debug bool, name string, command string, args ...string) (string, error) {
|
||||
assert.Equal(t, "tell", args[0])
|
||||
assert.Equal(t, "mds.a", args[1])
|
||||
assert.Equal(t, "version", args[2])
|
||||
return "", nil
|
||||
}
|
||||
context := &clusterd.Context{Executor: executor}
|
||||
|
||||
_, err := getCephDaemonVersionString(context, deployment, "rook-ceph")
|
||||
assert.Nil(t, err)
|
||||
}
|
||||
|
||||
@@ -55,7 +70,6 @@ func TestEnableMessenger2(t *testing.T) {
|
||||
executor.MockExecuteCommandWithOutput = func(debug bool, name string, command string, args ...string) (string, error) {
|
||||
assert.Equal(t, "mon", args[0])
|
||||
assert.Equal(t, "enable-msgr2", args[1])
|
||||
assert.Equal(t, 2, len(args))
|
||||
return "", nil
|
||||
}
|
||||
context := &clusterd.Context{Executor: executor}
|
||||
@@ -69,8 +83,6 @@ func TestEnableNautilusOSD(t *testing.T) {
|
||||
executor.MockExecuteCommandWithOutput = func(debug bool, name string, command string, args ...string) (string, error) {
|
||||
assert.Equal(t, "osd", args[0])
|
||||
assert.Equal(t, "require-osd-release", args[1])
|
||||
assert.Equal(t, "nautilus", args[2])
|
||||
assert.Equal(t, 3, len(args))
|
||||
return "", nil
|
||||
}
|
||||
context := &clusterd.Context{Executor: executor}
|
||||
@@ -78,3 +90,93 @@ func TestEnableNautilusOSD(t *testing.T) {
|
||||
err := EnableNautilusOSD(context)
|
||||
assert.Nil(t, err)
|
||||
}
|
||||
|
||||
func TestOkToStopDaemon(t *testing.T) {
|
||||
executor := &exectest.MockExecutor{}
|
||||
executor.MockExecuteCommandWithOutput = func(debug bool, name string, command string, args ...string) (string, error) {
|
||||
switch {
|
||||
case args[0] == "create" && args[1] == "mon" && args[2] == "a":
|
||||
assert.Equal(t, 2, len(args))
|
||||
return "", nil
|
||||
}
|
||||
return "", fmt.Errorf("unexpected ceph command '%v'", args)
|
||||
}
|
||||
context := &clusterd.Context{Executor: executor}
|
||||
|
||||
deployment := "rook-ceph-mon-a"
|
||||
err := okToStopDaemon(context, deployment, "rook-ceph")
|
||||
assert.Nil(t, err)
|
||||
|
||||
deployment = "rook-ceph-mgr-a"
|
||||
err = okToStopDaemon(context, deployment, "rook-ceph")
|
||||
assert.Nil(t, err)
|
||||
|
||||
deployment = "rook-ceph-dummy-a"
|
||||
err = okToStopDaemon(context, deployment, "rook-ceph")
|
||||
assert.Nil(t, err)
|
||||
}
|
||||
|
||||
func TestFindDaemonName(t *testing.T) {
|
||||
n, err := findDaemonName("rook-ceph-mon-a")
|
||||
assert.Nil(t, err)
|
||||
assert.Equal(t, "mon", n)
|
||||
n, err = findDaemonName("rook-ceph-osd-0")
|
||||
assert.Nil(t, err)
|
||||
assert.Equal(t, "osd", n)
|
||||
n, err = findDaemonName("rook-ceph-rgw-my-store-a")
|
||||
assert.Nil(t, err)
|
||||
assert.Equal(t, "rgw", n)
|
||||
n, err = findDaemonName("rook-ceph-mds-myfs-a")
|
||||
assert.Nil(t, err)
|
||||
assert.Equal(t, "mds", n)
|
||||
n, err = findDaemonName("rook-ceph-mgr-a")
|
||||
assert.Nil(t, err)
|
||||
assert.Equal(t, "mgr", n)
|
||||
n, err = findDaemonName("rook-ceph-rbd-mirror-a")
|
||||
assert.Nil(t, err)
|
||||
assert.Equal(t, "rbd-mirror", n)
|
||||
_, err = findDaemonName("rook-ceph-unknown-a")
|
||||
assert.NotNil(t, err)
|
||||
}
|
||||
|
||||
func TestFindDaemonID(t *testing.T) {
|
||||
id := findDaemonID("rook-ceph-mon-a")
|
||||
assert.Equal(t, "a", id)
|
||||
id = findDaemonID("rook-ceph-osd-0")
|
||||
assert.Equal(t, "0", id)
|
||||
id = findDaemonID("rook-ceph-rgw-my-super-store-a")
|
||||
assert.Equal(t, "a", id)
|
||||
id = findDaemonID("rook-ceph-mds-my-wonderful-fs-a")
|
||||
assert.Equal(t, "a", id)
|
||||
id = findDaemonID("rook-ceph-mgr-a")
|
||||
assert.Equal(t, "a", id)
|
||||
id = findDaemonID("rook.ceph.mgr.a")
|
||||
assert.NotEqual(t, "a", id)
|
||||
}
|
||||
|
||||
func TestOkToContinue(t *testing.T) {
|
||||
executor := &exectest.MockExecutor{}
|
||||
context := &clusterd.Context{Executor: executor}
|
||||
|
||||
err := OkToContinue(context, "rook-ceph", "rook-ceph-mon-a") // mon is not checked on ok-to-continue so nil is expected
|
||||
assert.Nil(t, err)
|
||||
}
|
||||
|
||||
func TestOkToStop(t *testing.T) {
|
||||
executor := &exectest.MockExecutor{}
|
||||
context := &clusterd.Context{Executor: executor}
|
||||
v := cephver.Nautilus
|
||||
|
||||
err := OkToStop(context, "rook-ceph", "rook-ceph-mon-a", "rook-ceph", v)
|
||||
assert.Nil(t, err)
|
||||
|
||||
err = OkToStop(context, "rook-ceph", "rook-ceph-mds-a", "rook-ceph", v)
|
||||
assert.Nil(t, err)
|
||||
}
|
||||
|
||||
func TestFindFSName(t *testing.T) {
|
||||
fsName := findFSName("rook-ceph-mds-myfs-a", "rook-ceph")
|
||||
assert.Equal(t, "myfs", fsName)
|
||||
fsName = findFSName("rook-ceph-mds-my-super-fs-a", "rook-ceph")
|
||||
assert.Equal(t, "my-super-fs", fsName)
|
||||
}
|
||||
@@ -29,6 +29,7 @@ import (
|
||||
"github.com/rook/rook/pkg/clusterd"
|
||||
"github.com/rook/rook/pkg/daemon/ceph/client"
|
||||
cephconfig "github.com/rook/rook/pkg/daemon/ceph/config"
|
||||
"github.com/rook/rook/pkg/operator/ceph/cluster/mon"
|
||||
"github.com/rook/rook/pkg/operator/ceph/config"
|
||||
opspec "github.com/rook/rook/pkg/operator/ceph/spec"
|
||||
"github.com/rook/rook/pkg/operator/k8sutil"
|
||||
@@ -106,7 +107,7 @@ func New(
|
||||
}
|
||||
}
|
||||
|
||||
var updateDeploymentAndWait = k8sutil.UpdateDeploymentAndWait
|
||||
var updateDeploymentAndWait = mon.UpdateCephDeploymentAndWait
|
||||
|
||||
// Start begins the process of running a cluster of Ceph mgrs.
|
||||
func (c *Cluster) Start() error {
|
||||
@@ -151,7 +152,7 @@ func (c *Cluster) Start() error {
|
||||
return fmt.Errorf("failed to create mgr deployment %s. %+v", resourceName, err)
|
||||
}
|
||||
logger.Infof("deployment for mgr %s already exists. updating if needed", resourceName)
|
||||
if _, err := updateDeploymentAndWait(c.context, d, c.Namespace); err != nil {
|
||||
if err := updateDeploymentAndWait(c.context, d, c.Namespace, c.clusterInfo.Name, c.clusterInfo.CephVersion); err != nil {
|
||||
return fmt.Errorf("failed to update mgr deployment %s. %+v", resourceName, err)
|
||||
}
|
||||
}
|
||||
|
||||
@@ -219,12 +219,12 @@ func (c *Cluster) startMons(targetCount int) error {
|
||||
|
||||
// Enable Ceph messenger 2 protocol on Nautilus
|
||||
if c.clusterInfo.CephVersion.IsAtLeastNautilus() {
|
||||
v, err := client.GetCephMonVersion(c.context)
|
||||
v, err := client.GetCephMonVersion(c.context, c.clusterInfo.Name)
|
||||
if err != nil {
|
||||
return fmt.Errorf("failed to get ceph mon version. %+v", err)
|
||||
}
|
||||
if v.IsAtLeastNautilus() {
|
||||
versions, err := client.GetCephVersions(c.context)
|
||||
versions, err := client.GetAllCephDaemonVersions(c.context, c.clusterInfo.Name)
|
||||
if err != nil {
|
||||
return fmt.Errorf("failed to get ceph daemons versions. %+v", err)
|
||||
}
|
||||
@@ -518,9 +518,28 @@ func (c *Cluster) startDeployments(mons []*monConfig, requireAllInQuorum bool) e
|
||||
if err != nil {
|
||||
return fmt.Errorf("failed to create mon %s. %+v", mons[i].DaemonName, err)
|
||||
}
|
||||
// For the initial deployment (first creation) it's expected to not have all the monitors in quorum
|
||||
// However, in an event of an update, it's crucial to proceed monitors by monitors
|
||||
// At the end of the method we perform one last check where all the monitors must be in quorum
|
||||
requireAllInQuorum := false
|
||||
err = c.waitForMonsToJoin(mons, requireAllInQuorum)
|
||||
if err != nil {
|
||||
return fmt.Errorf("failed to check mon quorum %s. %+v", mons[i].DaemonName, err)
|
||||
}
|
||||
}
|
||||
|
||||
logger.Infof("mons created: %d", len(mons))
|
||||
// Final verification that **all** mons are in quorum
|
||||
// Do not proceed if one monitor is still syncing
|
||||
// Only do this when monitors versions are different so we don't block the orchestration if a mon is down.
|
||||
versions, err := client.GetAllCephDaemonVersions(c.context, c.clusterInfo.Name)
|
||||
if err != nil {
|
||||
logger.Warningf("failed to get ceph daemons versions; this likely means there is no cluster yet. %+v", err)
|
||||
} else {
|
||||
if len(versions.Mon) != 1 {
|
||||
requireAllInQuorum = true
|
||||
}
|
||||
}
|
||||
return c.waitForMonsToJoin(mons, requireAllInQuorum)
|
||||
}
|
||||
|
||||
@@ -596,9 +615,10 @@ func (c *Cluster) saveMonConfig() error {
|
||||
return nil
|
||||
}
|
||||
|
||||
var updateDeploymentAndWait = k8sutil.UpdateDeploymentAndWait
|
||||
var updateDeploymentAndWait = UpdateCephDeploymentAndWait
|
||||
|
||||
func (c *Cluster) startMon(m *monConfig, hostname string) error {
|
||||
|
||||
d := c.makeDeployment(m, hostname)
|
||||
logger.Debugf("Starting mon: %+v", d.Name)
|
||||
_, err := c.context.Clientset.AppsV1().Deployments(c.Namespace).Create(d)
|
||||
@@ -607,7 +627,7 @@ func (c *Cluster) startMon(m *monConfig, hostname string) error {
|
||||
return fmt.Errorf("failed to create mon deployment %s. %+v", m.ResourceName, err)
|
||||
}
|
||||
logger.Infof("deployment for mon %s already exists. updating if needed", m.ResourceName)
|
||||
if _, err := updateDeploymentAndWait(c.context, d, c.Namespace); err != nil {
|
||||
if err := updateDeploymentAndWait(c.context, d, c.Namespace, c.clusterInfo.Name, c.clusterInfo.CephVersion); err != nil {
|
||||
return fmt.Errorf("failed to update mon deployment %s. %+v", m.ResourceName, err)
|
||||
}
|
||||
}
|
||||
@@ -620,7 +640,7 @@ func waitForQuorumWithMons(context *clusterd.Context, clusterName string, mons [
|
||||
|
||||
// wait for monitors to establish quorum
|
||||
retryCount := 0
|
||||
retryMax := 20
|
||||
retryMax := 30
|
||||
for {
|
||||
retryCount++
|
||||
if retryCount > retryMax {
|
||||
|
||||
@@ -21,9 +21,11 @@ import (
|
||||
"os"
|
||||
|
||||
cephv1 "github.com/rook/rook/pkg/apis/ceph.rook.io/v1"
|
||||
"github.com/rook/rook/pkg/clusterd"
|
||||
"github.com/rook/rook/pkg/daemon/ceph/client"
|
||||
"github.com/rook/rook/pkg/operator/ceph/config"
|
||||
|
||||
opspec "github.com/rook/rook/pkg/operator/ceph/spec"
|
||||
cephver "github.com/rook/rook/pkg/operator/ceph/version"
|
||||
"github.com/rook/rook/pkg/operator/k8sutil"
|
||||
apps "k8s.io/api/apps/v1"
|
||||
v1 "k8s.io/api/core/v1"
|
||||
@@ -213,3 +215,29 @@ func (c *Cluster) makeMonDaemonContainer(monConfig *monConfig) v1.Container {
|
||||
|
||||
return container
|
||||
}
|
||||
|
||||
// UpdateCephDeploymentAndWait verifies a deployment can be stopped or continued
|
||||
func UpdateCephDeploymentAndWait(context *clusterd.Context, deployment *apps.Deployment, namespace, clusterName string, cephVersion cephver.CephVersion) error {
|
||||
callback := func(action string) error {
|
||||
logger.Infof("checking if we can %s the deployment %s", action, deployment.Name)
|
||||
|
||||
if action == "stop" {
|
||||
err := client.OkToStop(context, namespace, deployment.Name, clusterName, cephVersion)
|
||||
if err != nil {
|
||||
return fmt.Errorf("failed to check if we can %s the deployment %s: %+v", action, deployment.Name, err)
|
||||
}
|
||||
}
|
||||
|
||||
if action == "continue" {
|
||||
err := client.OkToContinue(context, namespace, deployment.Name)
|
||||
if err != nil {
|
||||
return fmt.Errorf("failed to check if we can %s the deployment %s: %+v", action, deployment.Name, err)
|
||||
}
|
||||
}
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
_, err := k8sutil.UpdateDeploymentAndWait(context, deployment, namespace, callback)
|
||||
return err
|
||||
}
|
||||
@@ -28,6 +28,7 @@ import (
|
||||
"github.com/rook/rook/pkg/clusterd"
|
||||
"github.com/rook/rook/pkg/daemon/ceph/client"
|
||||
cephconfig "github.com/rook/rook/pkg/daemon/ceph/config"
|
||||
"github.com/rook/rook/pkg/operator/ceph/cluster/mon"
|
||||
osdconfig "github.com/rook/rook/pkg/operator/ceph/cluster/osd/config"
|
||||
opspec "github.com/rook/rook/pkg/operator/ceph/spec"
|
||||
cephver "github.com/rook/rook/pkg/operator/ceph/version"
|
||||
@@ -40,7 +41,10 @@ import (
|
||||
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
|
||||
)
|
||||
|
||||
var logger = capnslog.NewPackageLogger("github.com/rook/rook", "op-osd")
|
||||
var (
|
||||
logger = capnslog.NewPackageLogger("github.com/rook/rook", "op-osd")
|
||||
updateDeploymentAndWait = mon.UpdateCephDeploymentAndWait
|
||||
)
|
||||
|
||||
const (
|
||||
appName = "rook-ceph-osd"
|
||||
@@ -191,9 +195,9 @@ func (c *Cluster) Start() error {
|
||||
// The following block is used to apply any command(s) required by an upgrade
|
||||
// The block below handles the upgrade from Mimic to Nautilus.
|
||||
if c.clusterInfo.CephVersion.IsAtLeastNautilus() {
|
||||
versions, err := client.GetCephVersions(c.context)
|
||||
versions, err := client.GetAllCephDaemonVersions(c.context, c.clusterInfo.Name)
|
||||
if err != nil {
|
||||
logger.Warningf("failed to get ceph daemons versions. this likely means there are no osds yet. %+v", err)
|
||||
logger.Warningf("failed to get ceph daemons versions; this likely means there are no osds yet. %+v", err)
|
||||
} else {
|
||||
// If length is one, this clearly indicates that all the osds are running the same version
|
||||
logger.Infof("len of version.Osd is %d", len(versions.Osd))
|
||||
@@ -321,7 +325,7 @@ func (c *Cluster) startOSDDaemonsOnNode(nodeName string, config *provisionConfig
|
||||
continue
|
||||
}
|
||||
logger.Infof("deployment for osd %d already exists. updating if needed", osd.ID)
|
||||
if _, err = k8sutil.UpdateDeploymentAndWait(c.context, dp, c.Namespace); err != nil {
|
||||
if err = updateDeploymentAndWait(c.context, dp, c.Namespace, c.clusterInfo.Name, c.clusterInfo.CephVersion); err != nil {
|
||||
config.addError(fmt.Sprintf("failed to update osd deployment %d. %+v", osd.ID, err))
|
||||
}
|
||||
}
|
||||
|
||||
@@ -25,6 +25,7 @@ import (
|
||||
rookalpha "github.com/rook/rook/pkg/apis/rook.io/v1alpha2"
|
||||
"github.com/rook/rook/pkg/clusterd"
|
||||
cephconfig "github.com/rook/rook/pkg/daemon/ceph/config"
|
||||
"github.com/rook/rook/pkg/operator/ceph/cluster/mon"
|
||||
"github.com/rook/rook/pkg/operator/ceph/config"
|
||||
opspec "github.com/rook/rook/pkg/operator/ceph/spec"
|
||||
"github.com/rook/rook/pkg/operator/k8sutil"
|
||||
@@ -87,7 +88,7 @@ func New(
|
||||
}
|
||||
}
|
||||
|
||||
var updateDeploymentAndWait = k8sutil.UpdateDeploymentAndWait
|
||||
var updateDeploymentAndWait = mon.UpdateCephDeploymentAndWait
|
||||
|
||||
// Start begins the process of running rbd mirroring daemons.
|
||||
func (m *Mirroring) Start() error {
|
||||
@@ -119,7 +120,7 @@ func (m *Mirroring) Start() error {
|
||||
return fmt.Errorf("failed to create %s deployment. %+v", resourceName, err)
|
||||
}
|
||||
logger.Infof("deployment for rbd-mirror %s already exists. updating if needed", resourceName)
|
||||
if _, err := updateDeploymentAndWait(m.context, d, m.Namespace); err != nil {
|
||||
if err := updateDeploymentAndWait(m.context, d, m.Namespace, m.ClusterInfo.Name, m.ClusterInfo.CephVersion); err != nil {
|
||||
// fail could be an issue updating label selector (immutable), so try del and recreate
|
||||
logger.Debugf("updateDeploymentAndWait failed for rbd-mirror %s. Attempting del-and-recreate. %+v", resourceName, err)
|
||||
err = m.context.Clientset.AppsV1().Deployments(m.Namespace).Delete(d.Name, &metav1.DeleteOptions{})
|
||||
|
||||
@@ -27,6 +27,7 @@ import (
|
||||
"github.com/rook/rook/pkg/clusterd"
|
||||
"github.com/rook/rook/pkg/daemon/ceph/client"
|
||||
cephconfig "github.com/rook/rook/pkg/daemon/ceph/config"
|
||||
"github.com/rook/rook/pkg/operator/ceph/cluster/mon"
|
||||
"github.com/rook/rook/pkg/operator/ceph/config"
|
||||
opspec "github.com/rook/rook/pkg/operator/ceph/spec"
|
||||
cephver "github.com/rook/rook/pkg/operator/ceph/version"
|
||||
@@ -94,7 +95,7 @@ func NewCluster(
|
||||
}
|
||||
|
||||
// UpdateDeploymentAndWait can be overridden for unit tests. Do not alter this for runtime operation.
|
||||
var UpdateDeploymentAndWait = k8sutil.UpdateDeploymentAndWait
|
||||
var UpdateDeploymentAndWait = mon.UpdateCephDeploymentAndWait
|
||||
|
||||
// Start starts or updates a Ceph mds cluster in Kubernetes.
|
||||
func (c *Cluster) Start() error {
|
||||
@@ -155,7 +156,7 @@ func (c *Cluster) Start() error {
|
||||
// keyring must be generated before update-and-wait since no keyring will prevent the
|
||||
// deployment from reaching ready state
|
||||
if createErr != nil && errors.IsAlreadyExists(createErr) {
|
||||
if _, err = UpdateDeploymentAndWait(c.context, d, c.fs.Namespace); err != nil {
|
||||
if err = UpdateDeploymentAndWait(c.context, d, c.fs.Namespace, c.clusterInfo.Name, c.clusterInfo.CephVersion); err != nil {
|
||||
return fmt.Errorf("failed to update mds deployment %s. %+v", d.Name, err)
|
||||
}
|
||||
}
|
||||
|
||||
@@ -36,7 +36,7 @@ const (
|
||||
ganeshaRadosGraceCmd = "ganesha-rados-grace"
|
||||
)
|
||||
|
||||
var updateDeploymentAndWait = k8sutil.UpdateDeploymentAndWait
|
||||
var updateDeploymentAndWait = opmon.UpdateCephDeploymentAndWait
|
||||
|
||||
// Create the ganesha server
|
||||
func (c *CephNFSController) upCephNFS(n cephv1.CephNFS, oldActive int) error {
|
||||
@@ -68,7 +68,7 @@ func (c *CephNFSController) upCephNFS(n cephv1.CephNFS, oldActive int) error {
|
||||
return fmt.Errorf("failed to create ganesha deployment. %+v", err)
|
||||
}
|
||||
logger.Infof("ganesha deployment %s already exists. updating if needed", deployment.Name)
|
||||
if _, err := updateDeploymentAndWait(c.context, deployment, n.Namespace); err != nil {
|
||||
if err := updateDeploymentAndWait(c.context, deployment, n.Namespace, c.clusterInfo.Name, c.clusterInfo.CephVersion); err != nil {
|
||||
return fmt.Errorf("failed to update ganesha deployment %s. %+v", deployment.Name, err)
|
||||
}
|
||||
} else {
|
||||
|
||||
@@ -131,7 +131,13 @@ func (c *Cluster) Start(rookImage string) error {
|
||||
return fmt.Errorf("failed to create %s deployment. %+v", appName, err)
|
||||
}
|
||||
logger.Infof("deployment for mgr %s already exists. updating if needed", appName)
|
||||
if _, err := k8sutil.UpdateDeploymentAndWait(c.context, deployment, c.Namespace); err != nil {
|
||||
|
||||
// placeholder for a verify callback
|
||||
// see comments on k8sutil.UpdateDeploymentAndWait's definition to understand its purpose
|
||||
callback := func(action string) error {
|
||||
return nil
|
||||
}
|
||||
if _, err := k8sutil.UpdateDeploymentAndWait(c.context, deployment, c.Namespace, callback); err != nil {
|
||||
return fmt.Errorf("failed to update mgr deployment %s. %+v", appName, err)
|
||||
}
|
||||
} else {
|
||||
|
||||
@@ -21,8 +21,9 @@ import (
|
||||
"time"
|
||||
|
||||
"github.com/rook/rook/pkg/clusterd"
|
||||
"k8s.io/api/apps/v1"
|
||||
"github.com/rook/rook/pkg/util"
|
||||
apps "k8s.io/api/apps/v1"
|
||||
v1 "k8s.io/api/apps/v1"
|
||||
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
|
||||
"k8s.io/client-go/kubernetes"
|
||||
)
|
||||
@@ -36,6 +37,7 @@ func GetDeploymentImage(clientset kubernetes.Interface, namespace, name, contain
|
||||
return GetDeploymentSpecImage(clientset, *d, container, false)
|
||||
}
|
||||
|
||||
// GetDeploymentSpecImage returns the image name from the spec
|
||||
func GetDeploymentSpecImage(clientset kubernetes.Interface, d apps.Deployment, container string, initContainer bool) (string, error) {
|
||||
image, err := GetSpecContainerImage(d.Spec.Template.Spec, container, initContainer)
|
||||
if err != nil {
|
||||
@@ -47,12 +49,27 @@ func GetDeploymentSpecImage(clientset kubernetes.Interface, d apps.Deployment, c
|
||||
|
||||
// UpdateDeploymentAndWait updates a deployment and waits until it is running to return. It will
|
||||
// error if the deployment does not exist to be updated or if it takes too long.
|
||||
func UpdateDeploymentAndWait(context *clusterd.Context, deployment *apps.Deployment, namespace string) (*v1.Deployment, error) {
|
||||
// This method has a generic callback function that each backend can rely on
|
||||
// It serves two purposes:
|
||||
// 1. verify that a resource can be stopped
|
||||
// 2. verify that we can continue the update procedure
|
||||
// Basically, we go one resource by one and check if we can stop and then if the ressource has been successfully updated
|
||||
// we check if we can go ahead and move to the next one.
|
||||
func UpdateDeploymentAndWait(context *clusterd.Context, deployment *apps.Deployment, namespace string, verifyCallback func(action string) error) (*v1.Deployment, error) {
|
||||
original, err := context.Clientset.AppsV1().Deployments(namespace).Get(deployment.Name, metav1.GetOptions{})
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("failed to get deployment %s. %+v", deployment.Name, err)
|
||||
}
|
||||
|
||||
// Let's verify the deployment can be stopped
|
||||
// retry for 5 times, every minute
|
||||
err = util.Retry(5, 60*time.Second, func() error {
|
||||
return verifyCallback("stop")
|
||||
})
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("failed to check if deployment %s can be updated: %+v", deployment.Name, err)
|
||||
}
|
||||
|
||||
logger.Infof("updating deployment %s", deployment.Name)
|
||||
if _, err := context.Clientset.AppsV1().Deployments(namespace).Update(deployment); err != nil {
|
||||
return nil, fmt.Errorf("failed to update deployment %s. %+v", deployment.Name, err)
|
||||
@@ -73,9 +90,15 @@ func UpdateDeploymentAndWait(context *clusterd.Context, deployment *apps.Deploym
|
||||
}
|
||||
if d.Status.ObservedGeneration != original.Status.ObservedGeneration && d.Status.UpdatedReplicas > 0 && d.Status.ReadyReplicas > 0 {
|
||||
logger.Infof("finished waiting for updated deployment %s", d.Name)
|
||||
|
||||
// Now we check if we can go to the next daemon
|
||||
err = verifyCallback("continue")
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("failed to check if deployment %s can be updated: %+v", deployment.Name, err)
|
||||
}
|
||||
|
||||
return d, nil
|
||||
}
|
||||
|
||||
logger.Debugf("deployment %s status=%+v", d.Name, d.Status)
|
||||
time.Sleep(time.Duration(sleepTime) * time.Second)
|
||||
}
|
||||
|
||||
@@ -2,8 +2,8 @@ package test
|
||||
|
||||
import (
|
||||
"github.com/rook/rook/pkg/clusterd"
|
||||
"github.com/rook/rook/pkg/operator/ceph/version"
|
||||
apps "k8s.io/api/apps/v1"
|
||||
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
|
||||
)
|
||||
|
||||
// UpdateDeploymentAndWaitStub returns a stub replacement for the UpdateDeploymentAndWait function
|
||||
@@ -15,13 +15,13 @@ import (
|
||||
// returns a pointer to this slice which the calling func may use to verify the expected contents of
|
||||
// deploymentsUpdated based on expected behavior.
|
||||
func UpdateDeploymentAndWaitStub() (
|
||||
stubFunc func(context *clusterd.Context, deployment *apps.Deployment, namespace string) (*apps.Deployment, error),
|
||||
stubFunc func(context *clusterd.Context, deployment *apps.Deployment, namespace, clusterName string, cephVersion version.CephVersion) error,
|
||||
deploymentsUpdated *[]*apps.Deployment,
|
||||
) {
|
||||
deploymentsUpdated = &[]*apps.Deployment{}
|
||||
stubFunc = func(context *clusterd.Context, deployment *apps.Deployment, namespace string) (*apps.Deployment, error) {
|
||||
stubFunc = func(context *clusterd.Context, deployment *apps.Deployment, namespace, clusterName string, cephVersion version.CephVersion) error {
|
||||
*deploymentsUpdated = append(*deploymentsUpdated, deployment)
|
||||
return &apps.Deployment{ObjectMeta: metav1.ObjectMeta{UID: "stub-deployment-uid"}}, nil
|
||||
return nil
|
||||
}
|
||||
return stubFunc, deploymentsUpdated
|
||||
}
|
||||
|
||||
Reference in new issue
Block a user