Progress toward issue #2003. Includes design from design doc PR #1578 Use init containers to create configuration for Ceph mgrs. There is only one init container in this design. The init container calls the Rook binary to create Ceph config files which are then shared with the mds daemon main container. Once this init is run, the main mds daemon is run. Leaving room to use the Ceph-versioned image in the future, call `ceph-mds --foreground ...` to run the Ceph mds. The refactor to using an init container also necessitated refactoring the mdses replicaset implementation to a deployment-per-pod implementation due to a chicken-egg problem. With a single container (in the before times) the Rook binary was able to call the ceph-mds daemon with an id generated from the pod name. Since the pod name is not known before runtime, and the id is one of the few params that must be specified to Ceph daemons on run, it is necessary to know the id beforehand; thus the move to a deployment architecture following the likes of the mon and mgr daemons. Signed-off-by: Blaine Gardner <blaine.gardner@suse.com>
267 lines
7.9 KiB
Go
267 lines
7.9 KiB
Go
/*
|
|
Copyright 2016 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 (
|
|
"fmt"
|
|
"net"
|
|
"os"
|
|
"path"
|
|
|
|
mondaemon "github.com/rook/rook/pkg/daemon/ceph/mon"
|
|
opspec "github.com/rook/rook/pkg/operator/ceph/spec"
|
|
"github.com/rook/rook/pkg/operator/k8sutil"
|
|
"k8s.io/api/core/v1"
|
|
extensions "k8s.io/api/extensions/v1beta1"
|
|
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
|
|
"k8s.io/kubernetes/pkg/kubelet/apis"
|
|
)
|
|
|
|
const (
|
|
// Full path of command used to invoke the monmap tool
|
|
monmaptoolCommand = "/usr/bin/monmaptool"
|
|
// Full path of the command used to invoke the Ceph mon daemon
|
|
cephMonCommand = "ceph-mon"
|
|
|
|
monmapFile = "monmap"
|
|
)
|
|
|
|
func (c *Cluster) getLabels(daemonName string) map[string]string {
|
|
// Mons have a service for each mon, so the additional pod data is relevant for its services
|
|
// Use pod labels to keep "mon: id" for legacy
|
|
labels := opspec.PodLabels(appName, c.Namespace, "mon", daemonName)
|
|
// Add "mon_cluster: <namespace>" for legacy
|
|
labels[monClusterAttr] = c.Namespace
|
|
return labels
|
|
}
|
|
|
|
func (c *Cluster) makeDeployment(monConfig *monConfig, hostname string) *extensions.Deployment {
|
|
d := &extensions.Deployment{
|
|
ObjectMeta: metav1.ObjectMeta{
|
|
Name: monConfig.ResourceName,
|
|
Namespace: c.Namespace,
|
|
Labels: c.getLabels(monConfig.DaemonName),
|
|
},
|
|
}
|
|
k8sutil.SetOwnerRef(c.context.Clientset, c.Namespace, &d.ObjectMeta, &c.ownerRef)
|
|
|
|
pod := c.makeMonPod(monConfig, hostname)
|
|
replicaCount := int32(1)
|
|
d.Spec = extensions.DeploymentSpec{
|
|
Template: v1.PodTemplateSpec{
|
|
ObjectMeta: pod.ObjectMeta,
|
|
Spec: pod.Spec,
|
|
},
|
|
Replicas: &replicaCount,
|
|
Strategy: extensions.DeploymentStrategy{
|
|
Type: extensions.RecreateDeploymentStrategyType,
|
|
},
|
|
}
|
|
|
|
return d
|
|
}
|
|
|
|
/*
|
|
* Pod spec
|
|
*/
|
|
|
|
func (c *Cluster) makeMonPod(monConfig *monConfig, hostname string) *v1.Pod {
|
|
podSpec := v1.PodSpec{
|
|
InitContainers: []v1.Container{
|
|
// Config file init performed by Rook
|
|
c.makeConfigInitContainer(monConfig),
|
|
// Ceph monmap init performed by 'monmaptool'
|
|
c.makeMonmapInitContainer(monConfig),
|
|
// mon filesystem init performed by mon daemon
|
|
c.makeMonFSInitContainer(monConfig),
|
|
},
|
|
Containers: []v1.Container{
|
|
c.makeMonDaemonContainer(monConfig),
|
|
},
|
|
RestartPolicy: v1.RestartPolicyAlways,
|
|
NodeSelector: map[string]string{apis.LabelHostname: hostname},
|
|
Volumes: opspec.PodVolumes(c.dataDirHostPath),
|
|
HostNetwork: c.HostNetwork,
|
|
}
|
|
if c.HostNetwork {
|
|
podSpec.DNSPolicy = v1.DNSClusterFirstWithHostNet
|
|
}
|
|
c.placement.ApplyToPodSpec(&podSpec)
|
|
// remove Pod (anti-)affinity because we have our own placement logic
|
|
c.placement.PodAffinity = nil
|
|
c.placement.PodAntiAffinity = nil
|
|
|
|
pod := &v1.Pod{
|
|
ObjectMeta: metav1.ObjectMeta{
|
|
Name: monConfig.ResourceName,
|
|
Namespace: c.Namespace,
|
|
Labels: c.getLabels(monConfig.DaemonName),
|
|
Annotations: map[string]string{},
|
|
},
|
|
Spec: podSpec,
|
|
}
|
|
|
|
return pod
|
|
}
|
|
|
|
/*
|
|
* Container specs
|
|
*/
|
|
|
|
// Init and daemon containers require the same context, so we call it 'pod' context
|
|
func podSecurityContext() *v1.SecurityContext {
|
|
privileged := false
|
|
if os.Getenv("ROOK_HOSTPATH_REQUIRES_PRIVILEGED") == "true" {
|
|
privileged = true
|
|
}
|
|
return &v1.SecurityContext{Privileged: &privileged}
|
|
}
|
|
|
|
func (c *Cluster) makeConfigInitContainer(monConfig *monConfig) v1.Container {
|
|
return v1.Container{
|
|
Name: opspec.ConfigInitContainerName,
|
|
Args: []string{
|
|
"ceph",
|
|
mondaemon.InitCommand,
|
|
fmt.Sprintf("--config-dir=%s", k8sutil.DataDir),
|
|
fmt.Sprintf("--name=%s", monConfig.DaemonName),
|
|
fmt.Sprintf("--port=%d", monConfig.Port),
|
|
fmt.Sprintf("--fsid=%s", c.clusterInfo.FSID),
|
|
},
|
|
Image: k8sutil.MakeRookImage(c.Version),
|
|
Env: []v1.EnvVar{
|
|
k8sutil.PodIPEnvVar(k8sutil.PrivateIPEnvVar),
|
|
{Name: k8sutil.PublicIPEnvVar, Value: monConfig.PublicIP},
|
|
ClusterNameEnvVar(c.Namespace),
|
|
EndpointEnvVar(),
|
|
SecretEnvVar(),
|
|
AdminSecretEnvVar(),
|
|
k8sutil.ConfigOverrideEnvVar(),
|
|
},
|
|
VolumeMounts: opspec.RookVolumeMounts(),
|
|
SecurityContext: podSecurityContext(),
|
|
Resources: c.resources,
|
|
}
|
|
}
|
|
|
|
func (c *Cluster) monmapFilePath(monConfig *monConfig) string {
|
|
return path.Join(
|
|
mondaemon.GetMonRunDirPath(c.context.ConfigDir, monConfig.DaemonName),
|
|
monmapFile,
|
|
)
|
|
}
|
|
|
|
func (c *Cluster) makeMonmapInitContainer(monConfig *monConfig) v1.Container {
|
|
// Add mons w/ monmaptool w/ args: [--add <mon-name> <mon-endpoint>]...
|
|
monmapAddMonArgs := []string{}
|
|
for _, mon := range c.clusterInfo.Monitors {
|
|
monmapAddMonArgs = append(monmapAddMonArgs, "--add", mon.Name, mon.Endpoint)
|
|
}
|
|
|
|
return v1.Container{
|
|
Name: "monmap-init",
|
|
Command: []string{
|
|
monmaptoolCommand,
|
|
},
|
|
Args: append(
|
|
[]string{
|
|
c.monmapFilePath(monConfig),
|
|
"--create",
|
|
"--clobber",
|
|
"--fsid", c.clusterInfo.FSID,
|
|
},
|
|
monmapAddMonArgs...,
|
|
),
|
|
Image: k8sutil.MakeRookImage(c.Version), // TODO: ceph:<vers> image
|
|
VolumeMounts: opspec.CephVolumeMounts(),
|
|
SecurityContext: podSecurityContext(),
|
|
// monmap creation does not require ports to be exposed
|
|
Resources: c.resources,
|
|
}
|
|
}
|
|
|
|
// args needed for all ceph-mon calls
|
|
func (c *Cluster) cephMonCommonArgs(monConfig *monConfig) []string {
|
|
return []string{
|
|
"--name", fmt.Sprintf("mon.%s", monConfig.DaemonName),
|
|
"--mon-data", mondaemon.GetMonDataDirPath(c.context.ConfigDir, monConfig.DaemonName),
|
|
}
|
|
}
|
|
|
|
func (c *Cluster) makeMonFSInitContainer(monConfig *monConfig) v1.Container {
|
|
return v1.Container{
|
|
Name: "mon-fs-init",
|
|
Command: []string{
|
|
cephMonCommand,
|
|
},
|
|
Args: append(
|
|
[]string{
|
|
"--mkfs",
|
|
"--monmap", c.monmapFilePath(monConfig),
|
|
},
|
|
c.cephMonCommonArgs(monConfig)...,
|
|
),
|
|
Image: k8sutil.MakeRookImage(c.Version), // TODO: ceph:<vers> image
|
|
VolumeMounts: opspec.CephVolumeMounts(),
|
|
SecurityContext: podSecurityContext(),
|
|
// filesystem creation does not require ports to be exposed
|
|
Resources: c.resources,
|
|
}
|
|
}
|
|
|
|
func (c *Cluster) makeMonDaemonContainer(monConfig *monConfig) v1.Container {
|
|
return v1.Container{
|
|
// The operator has set up the mon's service already, so the IP that the mon should
|
|
// broadcast as its own (--public-addr) is known. But the pod's IP, which the mon should
|
|
// bind to (--public-bind-addr) isn't known until runtime. 3 solutions were considered for
|
|
// resolving this issue:
|
|
// 1. Rook in config init sets "public_bind_addr" in the Ceph config file
|
|
// - Chosen solution, but is not as transparent to inspection as using commandline arg
|
|
// 2. Use bash to do variable substitution with the pod IP env var; but bash is a poor PID1
|
|
// 3. Use tini to do var substitution as above; but tini doesn't exist in the ceph images.
|
|
Name: "mon",
|
|
Command: []string{
|
|
cephMonCommand,
|
|
},
|
|
Args: append(
|
|
[]string{
|
|
"--foreground",
|
|
"--public-addr", joinHostPort(monConfig.PublicIP, monConfig.Port),
|
|
// --public-bind-addr is set in the config file at init time
|
|
// do not add the '--cluster/--conf/--keyring' flags; rook wants their default values
|
|
},
|
|
c.cephMonCommonArgs(monConfig)...,
|
|
),
|
|
Image: k8sutil.MakeRookImage(c.Version), // TODO: ceph:<vers> image
|
|
VolumeMounts: opspec.CephVolumeMounts(),
|
|
SecurityContext: podSecurityContext(),
|
|
Ports: []v1.ContainerPort{
|
|
{
|
|
Name: "client",
|
|
ContainerPort: monConfig.Port,
|
|
Protocol: v1.ProtocolTCP,
|
|
},
|
|
},
|
|
Env: k8sutil.ClusterDaemonEnvVars(),
|
|
Resources: c.resources,
|
|
}
|
|
}
|
|
|
|
func joinHostPort(host string, port int32) string {
|
|
return net.JoinHostPort(host, fmt.Sprintf("%d", port))
|
|
}
|