ceph-csi/cmd/cephcsi.go
Prasanna Kumar Kalever b6a88dd728 rbd: add volume healer
Problem:
-------
For rbd nbd userspace mounter backends, after a restart of the nodeplugin
all the mounts will start seeing IO errors. This is because, for rbd-nbd
backends there will be a userspace mount daemon running per volume, post
restart of the nodeplugin pod, there is no way to restore the daemons
back to life.

Solution:
--------
The volume healer is a one-time activity that is triggered at the startup
time of the rbd nodeplugin. It navigates through the list of volume
attachments on the node and acts accordingly.

For now, it is limited to nbd type storage only, but it is flexible and
can be extended in the future for other backend types as needed.

From a few feets above:
This solves a severe problem for nbd backed csi volumes. The healer while
going through the list of volume attachments on the node, if finds the
volume is in attached state and is of type nbd, then it will attempt to
fix the rbd-nbd volumes by sending a NodeStageVolume request with the
required volume attributes like secrets, device name, image attributes,
and etc.. which will finally help start the required rbd-nbd daemons in
the nodeplugin csi-rbdplugin container. This will allow reattaching the
backend images with the right nbd device, thus allowing the applications
to perform IO without any interruptions even after a nodeplugin restart.

Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
2021-07-16 16:30:58 +00:00

278 lines
8.6 KiB
Go

/*
Copyright 2019 The Ceph-CSI Authors.
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 main
import (
"flag"
"fmt"
"os"
"runtime"
"time"
"github.com/ceph/ceph-csi/internal/cephfs"
"github.com/ceph/ceph-csi/internal/controller"
"github.com/ceph/ceph-csi/internal/controller/persistentvolume"
"github.com/ceph/ceph-csi/internal/liveness"
"github.com/ceph/ceph-csi/internal/rbd"
"github.com/ceph/ceph-csi/internal/util"
"k8s.io/klog/v2"
)
const (
rbdType = "rbd"
cephfsType = "cephfs"
livenessType = "liveness"
controllerType = "controller"
rbdDefaultName = "rbd.csi.ceph.com"
cephfsDefaultName = "cephfs.csi.ceph.com"
livenessDefaultName = "liveness.csi.ceph.com"
pollTime = 60 // seconds
probeTimeout = 3 // seconds
// use default namespace if namespace is not set.
defaultNS = "default"
defaultPluginPath = "/var/lib/kubelet/plugins"
defaultStagingPath = defaultPluginPath + "/kubernetes.io/csi/pv/"
)
var conf util.Config
func init() {
// common flags
flag.StringVar(&conf.Vtype, "type", "", "driver type [rbd|cephfs|liveness|controller]")
flag.StringVar(&conf.Endpoint, "endpoint", "unix://tmp/csi.sock", "CSI endpoint")
flag.StringVar(&conf.DriverName, "drivername", "", "name of the driver")
flag.StringVar(&conf.DriverNamespace, "drivernamespace", defaultNS, "namespace in which driver is deployed")
flag.StringVar(&conf.NodeID, "nodeid", "", "node id")
flag.StringVar(&conf.PluginPath, "pluginpath", defaultPluginPath, "plugin path")
flag.StringVar(&conf.StagingPath, "stagingpath", defaultStagingPath, "staging path")
flag.StringVar(&conf.InstanceID, "instanceid", "", "Unique ID distinguishing this instance of Ceph CSI among other"+
" instances, when sharing Ceph clusters across CSI instances for provisioning")
flag.IntVar(&conf.PidLimit, "pidlimit", 0, "the PID limit to configure through cgroups")
flag.BoolVar(&conf.IsControllerServer, "controllerserver", false, "start cephcsi controller server")
flag.BoolVar(&conf.IsNodeServer, "nodeserver", false, "start cephcsi node server")
flag.StringVar(
&conf.DomainLabels,
"domainlabels",
"",
"list of kubernetes node labels, that determines the topology"+
" domain the node belongs to, separated by ','")
// cephfs related flags
flag.BoolVar(
&conf.ForceKernelCephFS,
"forcecephkernelclient",
false,
"enable Ceph Kernel clients on kernel < 4.17 which support quotas")
// liveness/grpc metrics related flags
flag.IntVar(&conf.MetricsPort, "metricsport", 8080, "TCP port for liveness/grpc metrics requests")
flag.StringVar(
&conf.MetricsPath,
"metricspath",
"/metrics",
"path of prometheus endpoint where metrics will be available")
flag.DurationVar(&conf.PollTime, "polltime", time.Second*pollTime, "time interval in seconds between each poll")
flag.DurationVar(&conf.PoolTimeout, "timeout", time.Second*probeTimeout, "probe timeout in seconds")
flag.BoolVar(&conf.EnableGRPCMetrics, "enablegrpcmetrics", false, "[DEPRECATED] enable grpc metrics")
flag.StringVar(
&conf.HistogramOption,
"histogramoption",
"0.5,2,6",
"[DEPRECATED] Histogram option for grpc metrics, should be comma separated value, "+
"ex:= 0.5,2,6 where start=0.5 factor=2, count=6")
flag.UintVar(
&conf.RbdHardMaxCloneDepth,
"rbdhardmaxclonedepth",
8,
"Hard limit for maximum number of nested volume clones that are taken before a flatten occurs")
flag.UintVar(
&conf.RbdSoftMaxCloneDepth,
"rbdsoftmaxclonedepth",
4,
"Soft limit for maximum number of nested volume clones that are taken before a flatten occurs")
flag.UintVar(
&conf.MaxSnapshotsOnImage,
"maxsnapshotsonimage",
450,
"Maximum number of snapshots allowed on rbd image without flattening")
flag.UintVar(
&conf.MinSnapshotsOnImage,
"minsnapshotsonimage",
250,
"Minimum number of snapshots required on rbd image to start flattening")
flag.BoolVar(&conf.SkipForceFlatten, "skipforceflatten", false,
"skip image flattening if kernel support mapping of rbd images which has the deep-flatten feature")
flag.BoolVar(&conf.Version, "version", false, "Print cephcsi version information")
flag.BoolVar(&conf.EnableProfiling, "enableprofiling", false, "enable go profiling")
klog.InitFlags(nil)
if err := flag.Set("logtostderr", "true"); err != nil {
klog.Exitf("failed to set logtostderr flag: %v", err)
}
flag.Parse()
}
func getDriverName() string {
// was explicitly passed a driver name
if conf.DriverName != "" {
return conf.DriverName
}
// select driver name based on volume type
switch conf.Vtype {
case rbdType:
return rbdDefaultName
case cephfsType:
return cephfsDefaultName
case livenessType:
return livenessDefaultName
default:
return ""
}
}
func main() {
if conf.Version {
fmt.Println("Cephcsi Version:", util.DriverVersion)
fmt.Println("Git Commit:", util.GitCommit)
fmt.Println("Go Version:", runtime.Version())
fmt.Println("Compiler:", runtime.Compiler)
fmt.Printf("Platform: %s/%s\n", runtime.GOOS, runtime.GOARCH)
if kv, err := util.GetKernelVersion(); err == nil {
fmt.Println("Kernel:", kv)
}
os.Exit(0)
}
util.DefaultLog("Driver version: %s and Git version: %s", util.DriverVersion, util.GitCommit)
if conf.Vtype == "" {
logAndExit("driver type not specified")
}
dname := getDriverName()
err := util.ValidateDriverName(dname)
if err != nil {
logAndExit(err.Error())
}
// the driver may need a higher PID limit for handling all concurrent requests
if conf.PidLimit != 0 {
currentLimit, pidErr := util.GetPIDLimit()
if pidErr != nil {
klog.Errorf("Failed to get the PID limit, can not reconfigure: %v", pidErr)
} else {
util.DefaultLog("Initial PID limit is set to %d", currentLimit)
err = util.SetPIDLimit(conf.PidLimit)
switch {
case err != nil:
klog.Errorf("Failed to set new PID limit to %d: %v", conf.PidLimit, err)
case conf.PidLimit == -1:
util.DefaultLog("Reconfigured PID limit to %d (max)", conf.PidLimit)
default:
util.DefaultLog("Reconfigured PID limit to %d", conf.PidLimit)
}
}
}
if conf.EnableGRPCMetrics || conf.Vtype == livenessType {
// validate metrics endpoint
conf.MetricsIP = os.Getenv("POD_IP")
if conf.MetricsIP == "" {
klog.Warning("missing POD_IP env var defaulting to 0.0.0.0")
conf.MetricsIP = "0.0.0.0"
}
err = util.ValidateURL(&conf)
if err != nil {
logAndExit(err.Error())
}
}
util.DefaultLog("Starting driver type: %v with name: %v", conf.Vtype, dname)
switch conf.Vtype {
case rbdType:
validateCloneDepthFlag(&conf)
validateMaxSnaphostFlag(&conf)
driver := rbd.NewDriver()
driver.Run(&conf)
case cephfsType:
driver := cephfs.NewDriver()
driver.Run(&conf)
case livenessType:
liveness.Run(&conf)
case controllerType:
cfg := controller.Config{
DriverName: dname,
Namespace: conf.DriverNamespace,
}
// initialize all controllers before starting.
initControllers()
err = controller.Start(cfg)
if err != nil {
logAndExit(err.Error())
}
}
os.Exit(0)
}
// initControllers will initialize all the controllers.
func initControllers() {
// Add list of controller here.
persistentvolume.Init()
}
func validateCloneDepthFlag(conf *util.Config) {
// keeping hardlimit to 14 as max to avoid max image depth
if conf.RbdHardMaxCloneDepth == 0 || conf.RbdHardMaxCloneDepth > 14 {
logAndExit("rbdhardmaxclonedepth flag value should be between 1 and 14")
}
if conf.RbdSoftMaxCloneDepth > conf.RbdHardMaxCloneDepth {
logAndExit("rbdsoftmaxclonedepth flag value should not be greater than rbdhardmaxclonedepth")
}
}
func validateMaxSnaphostFlag(conf *util.Config) {
// maximum number of snapshots on an image are 510 [1] and 16 images in
// a parent/child chain [2],keeping snapshot limit to 500 to avoid issues.
// [1] https://github.com/torvalds/linux/blob/master/drivers/block/rbd.c#L98
// [2] https://github.com/torvalds/linux/blob/master/drivers/block/rbd.c#L92
if conf.MaxSnapshotsOnImage == 0 || conf.MaxSnapshotsOnImage > 500 {
logAndExit("maxsnapshotsonimage flag value should be between 1 and 500")
}
if conf.MinSnapshotsOnImage > conf.MaxSnapshotsOnImage {
logAndExit("minsnapshotsonimage flag value should be less than maxsnapshotsonimage")
}
}
func logAndExit(msg string) {
klog.Errorln(msg)
os.Exit(1)
}