容灾恢复 | 记一次K8S集群中etcd数据快照的备份恢复实践
设为「️ 星标」带你从基础入门 到 全栈实践 再到 放弃学习!
涉及 网络安全运维、应用开发、物联网IOT、学习路径 、个人感悟 等知识分享。
希望各位看友多多支持【关注、点赞、评论、收藏、投币】,助力每一个梦想。
【WeiyiGeek Blog's - 花开堪折直须折,莫待无花空折枝 】
作者主页: 【 https://weiyigeek.top 】
博客地址: 【 https://blog.weiyigeek.top 】
作者答疑学习交流群:欢迎各位志同道合的朋友一起学习交流【点击 加入交流群】, 或者关注公众号回复【学习交流群】。
0x00 前言简述
描述:在 Kubernetes 集群中所有操作的资源数据都是存储在 etcd 数据库上, 所以防止集群节点瘫痪未正常工作或在集群迁移时,以及在出现异常的情况下能尽快的恢复集群数据,则我们需要定期针对etcd集群数据进行相应的容灾操作。
在K8S集群中或者Docker环境中,我们可以非常方便的针对 etcd 数据进行备份,通我们常在一个节点上对 etcd 做快照就能够实现etcd数据的备份,其快照文件包含所有 Kubernetes 状态和关键信息, 有了etcd集群数据备份后,例如在灾难场景(例如丢失所有控制平面节点)下也能快速恢复 Kubernetes 集群,Boss再也不同担心系统起不来呢。
0x01 环境准备
1.安装的二进制 etcdctl
描述: etcdctl 二进制文件可以在 github.com/coreos/etcd/releases
选择对应的版本下载,例如可以执行以下 install_etcdctl.sh
的脚本,修改其中的版本信息。
install_etcdctl.sh
#!/bin/bash
# Author: WeiyiGeek
# Description: etcd 与 etcdctl 下载安装
ETCD_VER=v3.5.5
ETCD_DIR=etcd-download
DOWNLOAD_URL=https://github.com/coreos/etcd/releases/download
# Download
mkdir ${ETCD_DIR}
cd ${ETCD_DIR}
wget ${DOWNLOAD_URL}/${ETCD_VER}/etcd-${ETCD_VER}-linux-amd64.tar.gz
tar -xzvf etcd-${ETCD_VER}-linux-amd64.tar.gz
# Install
cd etcd-${ETCD_VER}-linux-amd64
cp etcdctl /usr/local/bin/
验证安装:
$ etcdctl version
etcdctl version: 3.5.5
API version: 3.5
2.拉取带有 etcdctl 的 Docker 镜像
操作流程:
# 镜像拉取与容器创建。
docker run --rm \
-v /data/backup:/backup \
-v /etc/kubernetes/pki/etcd:/etc/kubernetes/pki/etcd \
--env ETCDCTL_API=3 \
registry.cn-hangzhou.aliyuncs.com/google_containers/etcd:3.5.1-0 \
/bin/sh -c "etcdctl version"
安装验证:
3.5.1-0: Pulling from google_containers/etcd
e8614d09b7be: Pull complete
45b6afb4a92f: Pull complete
.......
Digest: sha256:64b9ea357325d5db9f8a723dcf503b5a449177b17ac87d69481e126bb724c263
Status: Downloaded newer image for registry.cn-hangzhou.aliyuncs.com/google_containers/etcd:3.5.1-0
etcdctl version: 3.5.1
API version: 3.5
3.Kubernetes 使用 etcdctl 镜像创建Pod
# 镜像拉取
crictl pull registry.cn-hangzhou.aliyuncs.com/google_containers/etcd:3.5.5-0
# Pod 创建以及安装验证
$ kubectl run etcdctl --image=registry.cn-hangzhou.aliyuncs.com/google_containers/etcd:3.5.5-0 --command -- /usr/local/bin/etcdctl version
$ kubectl logs -f etcdctl
etcdctl version: 3.5.5
API version: 3.5
$ kubectl delete pod etcdctl
0x02 备份实践
1.使用二进制安装 etcdctl 客户端工具
温馨提示: 如果是单节点 Kubernetes 我们只需要对其的 etcd 数据库进行快照备份, 如果是多主多从的集群,我们则需依次备份多个 master 节点中 etcd,防止在备份时etc数据被更改!
此处实践环境为多master高可用集群节点, 即三主节点、四从工作节点,若你对K8s集群不了解或者项搭建高可用集群的朋友,关注 WeiyiGeek 公众号回复【Kubernetes学习之路汇总
】即可获得学习资料:
https://www.weiyigeek.top/wechat.html?key=Kubernetes学习之路汇总
$ kubectl get node
NAME STATUS ROLES AGE VERSION
weiyigeek-107 Ready control-plane,master 279d v1.23.1
weiyigeek-108 Ready control-plane,master 278d v1.23.1
weiyigeek-109 Ready control-plane,master 278d v1.23.1
weiyigeek-223 Ready work 278d v1.23.1
weiyigeek-224 Ready work 278d v1.23.1
weiyigeek-225 Ready work 279d v1.23.1
weiyigeek-226 Ready work 118d v1.23.1
操作流程:
# 1.创建etcd快照备份目录
$ mkdir -pv /backup
# 2.查看etcd证书
$ ls /etc/kubernetes/pki/etcd/
ca.crt ca.key healthcheck-client.crt healthcheck-client.key peer.crt peer.key server.crt server.key
# 3.查看 etcd 地址以及服务
$ kubectl get pod -n kube-system -o wide | grep "etcd"
etcd-weiyigeek-107 1/1 Running 1 279d 192.168.12.107 weiyigeek-107 <none><none>
etcd-weiyigeek-108 1/1 Running 0 278d 192.168.12.108 weiyigeek-108 <none><none>
etcd-weiyigeek-109 1/1 Running 0 278d 192.168.12.109 weiyigeek-109 <none><none>
# 4.此时在 107、 108 、109 主节点上查看你监听情况
$ netstat -ano | grep "107:2379"
tcp 0 0 192.168.12.107:2379 0.0.0.0:* LISTEN off (0.00/0/0)
$ netstat -ano | grep "108:2379"
tcp 0 0 192.168.12.108:2379 0.0.0.0:* LISTEN off (0.00/0/0)
$ netstat -ano | grep "109:2379"
tcp 0 0 192.168.12.109:2379 0.0.0.0:* LISTEN off (0.00/0/0)
# 5. 使用etcdctl客户端工具依次备份节点中的数据
$ etcdctl --endpoints=https://10.20.176.212:2379 \
--cacert=/etc/kubernetes/pki/etcd/ca.crt \
--cert=/etc/kubernetes/pki/etcd/healthcheck-client.crt \
--key=/etc/kubernetes/pki/etcd/healthcheck-client.key \
snapshot save /backup/etcd-snapshot.db
{"level":"info","ts":"2022-10-23T16:32:26.020+0800","caller":"snapshot/v3_snapshot.go:65","msg":"created temporary db file","path":"/backup/etcd-snapshot.db.part"}
{"level":"info","ts":"2022-10-23T16:32:26.034+0800","logger":"client","caller":"v3/maintenance.go:211","msg":"opened snapshot stream; downloading"}
{"level":"info","ts":"2022-10-23T16:32:26.034+0800","caller":"snapshot/v3_snapshot.go:73","msg":"fetching snapshot","endpoint":"https://10.20.176.212:2379"}
{"level":"info","ts":"2022-10-23T16:32:26.871+0800","logger":"client","caller":"v3/maintenance.go:219","msg":"completed snapshot read; closing"}
{"level":"info","ts":"2022-10-23T16:32:26.946+0800","caller":"snapshot/v3_snapshot.go:88","msg":"fetched snapshot","endpoint":"https://10.20.176.212:2379","size":"112 MB","took":"now"}
{"level":"info","ts":"2022-10-23T16:32:26.946+0800","caller":"snapshot/v3_snapshot.go:97","msg":"saved","path":"/backup/etcd-snapshot.db"}
Snapshot saved at /backup/etcd-snapshot.db
通过etcdctl查询Kubernetes中etcd数据,由于Kubernetes使用etcd v3版本的API,而且etcd 集群中默认使用tls认证,所以先配置几个环境变量。
# 1.环境变量
export ETCDCTL_API=3
export ETCDCTL_CACERT=/etc/kubernetes/pki/etcd/ca.crt
export ETCDCTL_CERT=/etc/kubernetes/pki/etcd/peer.crt
export ETCDCTL_KEY=/etc/kubernetes/pki/etcd/peer.key
# 2.查询集群中所有的key列表
# –-prefix:表示查找所有以/registry为前缀的key
# --keys-only=true:表示只给出key不给出value
etcdctl --endpoints=https://10.20.176.212:2379 get /registry --prefix --keys-only=true | head -n 1
/registry/apiextensions.k8s.io/customresourcedefinitions/bgpconfigurations.crd.projectcalico.org
# 3.查询某个key的值
# -–keys-only=false : 表示要给出value,该参数默认值即为false,
# -w=json :表示输出json格式
etcdctl --endpoints=https://10.20.176.212:2379 get /registry/namespaces/default --prefix --keys-only=false -w=json | python3 -m json.tool
etcdctl --endpoints=https://10.20.176.212:2379 get /registry/namespaces/default --prefix --keys-only=false -w=json | python3 -m json.tool
{
"header": {
"cluster_id": 11404821125176160774,
"member_id": 7099450421952911102,
"revision": 30240109,
"raft_term": 3
},
"kvs": [
{
"key": "L3JlZ2lzdHJ5L25hbWVzcGFjZXMvZGVmYXVsdA==",
"create_revision": 192,
"mod_revision": 192,
"version": 1,
"value": "azhzAAoPCg.......XNwYWNlEGgAiAA=="
}
],
"count": 1
}
# 4.其 Key / Value 都是采用 base64 编码
echo -e "L3JlZ2lzdHJ5L25hbWVzcGFjZXMvZGVmYXVsdA==" | base64 -d
echo -e "azhzAAoPCgJ2MRIJTmFtZXNwYWNlEogCCu0BCgdkZWZhdWx0EgAaACIAKiQ5ZDQyYmYxMy03OGM0LTQ4NzQtOThiYy05NjNlMDg1MDYyZjYyADgAQggIo8yllQYQAFomChtrdWJlcm5ldGVzLmlvL21ldGFkYXRhLm5hbWUSB2RlZmF1bHR6AIoBfQoOa3ViZS1hcGlzZXJ2ZXISBlVwZGF0ZRoCdjEiCAijzKWVBhAAMghGaWVsZHNWMTpJCkd7ImY6bWV0YWRhdGEiOnsiZjpsYWJlbHMiOnsiLiI6e30sImY6a3ViZXJuZXRlcy5pby9tZXRhZGF0YS5uYW1lIjp7fX19fUIAEgwKCmt1YmVybmV0ZXMaCAoGQWN0aXZlGgAiAA==" | base64 -d
# 5.实际上述value编码解码后的的内容如下:
$ kubectl get ns default -o yaml
apiVersion: v1
kind: Namespace
metadata:
creationTimestamp: "2022-06-15T04:54:59Z"
labels:
kubernetes.io/metadata.name: default
name: default
resourceVersion: "192"
uid: 9d42bf13-78c4-4874-98bc-963e085062f6
spec:
finalizers:
- kubernetes
status:
phase: Active
# 5.Pod 资源信息查看
etcdctl --endpoints=https://10.20.176.212:2379 get /registry/pods/default --prefix --keys-only=true
# /registry/pods/default/nfs-dev-nfs-subdir-external-provisioner-cf7684f8b-fzl9h
# /registry/pods/default/nfs-local-nfs-subdir-external-provisioner-6f97d44bb8-424tk
etcdctl --endpoints=https://10.20.176.212:2379 get /registry/pods/default --prefix --keys-only=true -w=json | python3 -m json.tool
{
"header": {
"cluster_id": 11404821125176160774,
"member_id": 7099450421952911102,
"revision": 30442415,
"raft_term": 3
},
"kvs": [
{ # 实际上该编码是 /registry/pods/default/nfs-dev-nfs-subdir-external-provisioner-cf7684f8b-fzl9h
"key": "L3JlZ2lzdHJ5L3BvZHMvZGVmYXVsdC9uZnMtZGV2LW5mcy1zdWJkaXItZXh0ZXJuYWwtcHJvdmlzaW9uZXItY2Y3Njg0ZjhiLWZ6bDlo",
"create_revision": 5510865,
"mod_revision": 5510883,
"version": 5
},
{
"key": "L3JlZ2lzdHJ5L3BvZHMvZGVmYXVsdC9uZnMtbG9jYWwtbmZzLXN1YmRpci1leHRlcm5hbC1wcm92aXNpb25lci02Zjk3ZDQ0YmI4LTQyNHRr",
"create_revision": 5510967,
"mod_revision": 5510987,
"version": 5
}
],
"count": 2
}
2.使用Docker镜像安装 etcdctl 客户端工具
描述: 在装有Docker环境的机器,我们可以非常方便的备份K8s集群中的etcd数据库,此处我已经安装好了Docker,若有不了解Docker或者需要搭建Docker环境中童鞋。
# 1.Docker 实践环境
$ docker version
Client: Docker Engine - Community
Version 20.10.3
.......
Server: Docker Engine - Community
Engine:
Version: 20.10.3
# 2.etcd 备份文件存储的目录
$ mkdir -vp /backup
# 3.执行docker创建容器,在备份数据库后便删除该容器。
$ docker run --rm \
-v /backup:/backup \
-v /etc/kubernetes/pki/etcd:/etc/kubernetes/pki/etcd \
--env ETCDCTL_API=3 \
registry.cn-hangzhou.aliyuncs.com/google_containers/etcd:3.5.1-0 \
/bin/sh -c "etcdctl --endpoints=https://192.168.12.107:2379 \
--cacert=/etc/kubernetes/pki/etcd/ca.crt \
--key=/etc/kubernetes/pki/etcd/healthcheck-client.key \
--cert=/etc/kubernetes/pki/etcd/healthcheck-client.crt \
snapshot save /backup/etcd-snapshot.db"
# {"level":"info","ts":1666515535.63076,"caller":"snapshot/v3_snapshot.go:68","msg":"created temporary db file","path":"/backup/etcd-snapshot.db.part"}
# {"level":"info","ts":1666515535.6411893,"logger":"client","caller":"v3/maintenance.go:211","msg":"opened snapshot stream; downloading"}
# {"level":"info","ts":1666515535.6419039,"caller":"snapshot/v3_snapshot.go:76","msg":"fetching snapshot","endpoint":"https://192.168.12.107:2379"}
# {"level":"info","ts":1666515535.9170482,"logger":"client","caller":"v3/maintenance.go:219","msg":"completed snapshot read; closing"}
# {"level":"info","ts":1666515535.931862,"caller":"snapshot/v3_snapshot.go:91","msg":"fetched snapshot","endpoint":"https://192.168.12.107:2379","size":"9.0 MB","took":"now"}
# {"level":"info","ts":1666515535.9322069,"caller":"snapshot/v3_snapshot.go:100","msg":"saved","path":"/backup/etcd-snapshot.db"}
Snapshot saved at /backup/etcd-snapshot.db
# 4.查看备份的etcd快照文件
ls -alh /backup/etcd-snapshot.db
-rw------- 1 root root 8.6M Oct 23 16:58 /backup/etcd-snapshot.db
使用 Docker 容器查看 k8s 集群中的etcd数据库中的数据。
$ docker run --rm \
-v /backup:/backup \
-v /etc/kubernetes/pki/etcd:/etc/kubernetes/pki/etcd \
--env ETCDCTL_API=3 \
--env ETCDCTL_CACERT=/etc/kubernetes/pki/etcd/ca.crt \
--env ETCDCTL_CERT=/etc/kubernetes/pki/etcd/peer.crt \
--env ETCDCTL_KEY=/etc/kubernetes/pki/etcd/peer.key \
registry.cn-hangzhou.aliyuncs.com/google_containers/etcd:3.5.1-0 \
/bin/sh -c "etcdctl --endpoints=https://192.168.12.107:2379 get /registry/namespaces/default -w=json"
执行结果:
3.在kubernetes集群中快速创建pod进行手动备份
准备一个Pod资源清单并部署
cat <<EOF | kubectl apply -f -
apiVersion: v1
kind: Pod
metadata:
name: etcd-backup
labels:
tool: backup
spec:
containers:
- name: redis
image: registry.cn-hangzhou.aliyuncs.com/google_containers/etcd:3.5.5-0
imagePullPolicy: IfNotPresent
command:
- sh
- -c
- "etcd"
env:
- name: ETCDCTL_API
value: "3"
- name: ETCDCTL_CACERT
value: "/etc/kubernetes/pki/etcd/ca.crt"
- name: ETCDCTL_CERT
value: "/etc/kubernetes/pki/etcd/healthcheck-client.crt"
- name: ETCDCTL_KEY
value: "/etc/kubernetes/pki/etcd/healthcheck-client.key"
volumeMounts:
- name: "pki"
mountPath: "/etc/kubernetes"
- name: "backup"
mountPath: "/backup"
volumes:
- name: pki
hostPath:
path: "/etc/kubernetes" # 证书目录
type: "DirectoryOrCreate"
- name: "backup"
hostPath:
path: "/storage/dev/backup" # 数据备份目录
type: "DirectoryOrCreate"
restartPolicy: Never
nodeSelector:
node-role.kubernetes.io/master: "" # 绑定在主节点中
EOF
pod/etcd-backup created
进入到该Pod终端之中执行相应的备份命令
~$ kubectl exec -it etcd-backup sh
# 快照备份
sh-5.1# export RAND=$RANDOM
sh-5.1# etcdctl --endpoints=https://192.168.12.107:2379 snapshot save /backup/etcd-107-${RAND}-snapshot.db
sh-5.1# etcdctl --endpoints=https://192.168.12.108:2379 snapshot save /backup/etcd-108-${RAND}-snapshot.db
Snapshot saved at /backup/etcd-108-32616-snapshot.db
sh-5.1# etcdctl --endpoints=https://192.168.12.109:2379 snapshot save /backup/etcd-109-${RAND}-snapshot.db
# etcd 节点成员
sh-5.1# etcdctl member list --endpoints=https://192.168.12.107:2379 --endpoints=https://192.168.12.108:2379 --endpoints=https://192.168.12.109:2379
2db31a5d67ec1034, started, weiyigeek-108, https://192.168.12.108:2380, https://192.168.12.108:2379, false
42efe7cca897d765, started, weiyigeek-109, https://192.168.12.109:2380, https://192.168.12.109:2379, false
471323846709334f, started, weiyigeek-107, https://192.168.12.107:2380, https://192.168.12.107:2379, false
# etcd 节点健康信息
sh-5.1# etcdctl endpoint health --endpoints=https://192.168.12.107:2379 --endpoints=https://192.168.12.108:2379 --endpoints=https://192.168.12.109:2379
https://192.168.12.107:2379 is healthy: successfully committed proposal: took = 11.930331ms
https://192.168.12.109:2379 is healthy: successfully committed proposal: took = 11.930993ms
https://192.168.12.108:2379 is healthy: successfully committed proposal: took = 109.515933ms
# etcd 节点状态及空间占用信息
sh-5.1# etcdctl endpoint status --endpoints=https://192.168.12.107:2379 --endpoints=https://192.168.12.108:2379 --endpoints=https://192.168.12.109:2379
https://192.168.12.107:2379, 471323846709334f, 3.5.1, 9.2 MB, false, false, 4, 71464830, 71464830,
https://192.168.12.108:2379, 2db31a5d67ec1034, 3.5.1, 9.2 MB, false, false, 4, 71464830, 71464830,
https://192.168.12.109:2379, 42efe7cca897d765, 3.5.1, 9.2 MB, true, false, 4, 71464830, 71464830, # 此处为主
至此,手动备份etcd集群数据快照完毕!
4.在kubernetes集群中使用CronJob资源控制器进行定时备份
首先准备一个cronJob资源清单:
cat > etcd-database-backup.yaml <<'EOF'
apiVersion: batch/v1
kind: CronJob
metadata:
name: etcd-database-backup
annotations:
descript: "etcd数据库定时备份"
spec:
schedule: "*/5 * * * *" # 表示每5分钟运行一次
jobTemplate:
spec:
template:
spec:
containers:
- name: etcdctl
image: registry.cn-hangzhou.aliyuncs.com/google_containers/etcd:3.5.5-0
env:
- name: ETCDCTL_API
value: "3"
- name: ETCDCTL_CACERT
value: "/etc/kubernetes/pki/etcd/ca.crt"
- name: ETCDCTL_CERT
value: "/etc/kubernetes/pki/etcd/healthcheck-client.crt"
- name: ETCDCTL_KEY
value: "/etc/kubernetes/pki/etcd/healthcheck-client.key"
command:
- /bin/sh
- -c
- |
export RAND=$RANDOM
etcdctl --endpoints=https://192.168.12.107:2379 snapshot save /backup/etcd-107-${RAND}-snapshot.db
etcdctl --endpoints=https://192.168.12.108:2379 snapshot save /backup/etcd-108-${RAND}-snapshot.db
etcdctl --endpoints=https://192.168.12.109:2379 snapshot save /backup/etcd-109-${RAND}-snapshot.db
volumeMounts:
- name: "pki"
mountPath: "/etc/kubernetes"
- name: "backup"
mountPath: "/backup"
imagePullPolicy: IfNotPresent
volumes:
- name: "pki"
hostPath:
path: "/etc/kubernetes"
type: "DirectoryOrCreate"
- name: "backup"
hostPath:
path: "/storage/dev/backup" # 数据备份目录
type: "DirectoryOrCreate"
nodeSelector: # 将Pod绑定在主节点之中,否则只能将相关证书放在各个节点能访问的nfs共享存储中
node-role.kubernetes.io/master: ""
restartPolicy: Never
EOF
创建cronjob资源清单:
kubectl apply -f etcd-database-backup.yaml
# cronjob.batch/etcd-database-backup created
查看创建的cronjob资源及其集群etcd备份:
NAME SCHEDULE SUSPEND ACTIVE LAST SCHEDULE AGE
cronjob.batch/etcd-database-backup */5 * * * * False 0 21s 14m
NAME READY STATUS RESTARTS AGE
pod/etcd-database-backup-27776740-rhzkk 0/1 Completed 0 21s
查看定时Pod日志以及备份文件:
kubectl logs -f pod/etcd-database-backup-27776740-rhzkk
Snapshot saved at /backup/etcd-107-25615-snapshot.db
Snapshot saved at /backup/etcd-108-25615-snapshot.db
Snapshot saved at /backup/etcd-109-25615-snapshot.db
$ ls -lt # 显示最新备份的文件按照时间排序
total 25M
-rw------- 1 root root 8.6M Oct 24 21:12 etcd-107-25615-snapshot.db
-rw------- 1 root root 7.1M Oct 24 21:12 etcd-108-25615-snapshot.db
-rw------- 1 root root 8.8M Oct 24 21:12 etcd-109-25615-snapshot.db
至此集群中的etcd快照数据备份完毕!
0x02 恢复实践
1.单master节点恢复
描述: 当单master集群节点资源清单数据丢失时,我们可采用如下方式进行快速恢复数据。
操作流程:
温馨提示: 如果是单节点的k8S集群则使用如下命令恢复
mv /etc/kubernetes/manifests/ /etc/kubernetes/manifests-backup/
mv /var/lib/etcd /var/lib/etcd.bak
ETCDCTL_API=3 etcdctl snapshot restore /backup/etcd-212-32616-snapshot.db --data-dir=/var/lib/etcd/ --endpoints=https://10.20.176.212:2379 --cacert=/etc/kubernetes/pki/etcd/ca.crt --cert=/etc/kubernetes/pki/etcd/peer.crt --key=/etc/kubernetes/pki/etcd/peer.key
2.多master节点恢复
1.温馨提示,此处的集群etcd数据库是安装在Kubernetes集群之中的,并非外部独立安装部署的。
$ kubectl get pod -n kube-system etcd-devtest-master-212 etcd-devtest-master-213 etcd-devtest-master-214
NAME READY STATUS RESTARTS AGE
etcd-devtest-master-212 1/1 Running 3 134d
etcd-devtest-master-213 1/1 Running 0 134d
etcd-devtest-master-214 1/1 Running 0 134d
2.前面我们提到过,在进行恢复前需要查看 etcd 集群当前成员以及监控状态。
# etcd 集群成员列表
ETCDCTL_API=3 etcdctl member list --endpoints=https://10.20.176.212:2379 --cacert=/etc/kubernetes/pki/etcd/ca.crt --cert=/etc/kubernetes/pki/etcd/peer.crt --key=/etc/kubernetes/pki/etcd/peer.key
6286508b550016fe, started, devtest-master-212, https://10.20.176.212:2380, https://10.20.176.212:2379, false
9dd15f852caf8e05, started, devtest-master-214, https://10.20.176.214:2380, https://10.20.176.214:2379, false
e0f23bd90b7c7c0d, started, devtest-master-213, https://10.20.176.213:2380, https://10.20.176.213:2379, false
# etcd 集群节点状态查看主从节点
ETCDCTL_API=3 etcdctl endpoint status --endpoints=https://10.20.176.212:2379 --endpoints=https://10.20.176.213:2379 --endpoints=https://10.20.176.214:2379 --cacert=/etc/kubernetes/pki/etcd/ca.crt --cert=/etc/kubernetes/pki/etcd/peer.crt --key=/etc/kubernetes/pki/etcd/peer.key --write-out table
# etcd 集群节点健康信息筛选出不健康的节点
ETCDCTL_API=3 etcdctl endpoint health --endpoints=https://10.20.176.212:2379 --endpoints=https://10.20.176.213:2379 --endpoints=https://10.20.176.214:2379 --cacert=/etc/kubernetes/pki/etcd/ca.crt --cert=/etc/kubernetes/pki/etcd/peer.crt --key=/etc/kubernetes/pki/etcd/peer.key
https://10.20.176.212:2379 is healthy: successfully committed proposal: took = 14.686875ms
https://10.20.176.214:2379 is healthy: successfully committed proposal: took = 16.201187ms
https://10.20.176.213:2379 is healthy: successfully committed proposal: took = 18.962462ms
3.停掉所有Master机器的kube-apiserver和etcd ,然后在利用备份进行恢复该节点的etcd数据。
mv /etc/kubernetes/manifests/ /etc/kubernetes/manifests-backup/
# 在该节点上删除 /var/lib/etcd
mv /var/lib/etcd /var/lib/etcd.bak
# 利用快照进行恢复,注意etcd集群中可以用同一份snapshot恢复。
ETCDCTL_API=3 etcdctl snapshot restore /backup/etcd-212-32616-snapshot.db --data-dir=/var/lib/etcd --name=devtest-master-212 --cacert=/etc/kubernetes/pki/etcd/ca.crt --cert=/etc/kubernetes/pki/etcd/peer.crt --key=/etc/kubernetes/pki/etcd/peer.key --initial-cluster-token=etcd-cluster-0 --initial-cluster=devtest-master-212=https://10.20.176.212:2380,devtest-master-213=https://10.20.176.213:2380,devtest-master-214=https://10.20.176.214:2380 --initial-advertise-peer-urls=https://10.20.176.212:2380
温馨提示: 当节点加入控制平面 control-plane 后为 API Server、Controller Manager 和 Scheduler
生成静态Pod配置清单,主机上的kubelet服务会监视 /etc/kubernetes/manifests
目录中的配置清单的创建、变动和删除等状态变动,并根据变动完成Pod创建、更新或删除操作。因此,这两个阶段创建生成的各配置清单将会启动Master组件的相关Pod
4.然后启动 etcd 和 apiserver 并查看 pods是否恢复正常。
# 查看 etcd pod
$ kubectl get pod -n kube-system -l component=etcd
# 查看组件健康状态
$ kubectl get componentstatuses
Warning: v1 ComponentStatus is deprecated in v1.19+
NAME STATUS MESSAGE ERROR
scheduler Healthy ok
controller-manager Healthy ok
etcd-0 Healthy {"health":"true","reason":""}
3.K8S集群中etcd数据库节点数据不一致问题解决实践
描述: 在日常巡检业务时,发现etcd数据不一致,执行kubectl get pod -n xxx获取的信息资源不一样, etcdctl 直接查询了 etcd 集群状态和集群数据,返回结果显示 3 个节点状态都正常,且 RaftIndex 一致,观察 etcd 的日志也并未发现报错信息,唯一可疑的地方是 3 个节点的 dbsize 差别较大,得出的结论是 etcd 数据不一致。
$ ETCDCTL_API=3 etcdctl endpoint status --endpoints=https://192.168.12.108:2379 --endpoints=https://192.168.12.107:2379 --endpoints=https://192.168.12.109:2379 --cacert=/etc/kubernetes/pki/etcd/ca.crt --cert=/etc/kubernetes/pki/etcd/peer.crt --key=/etc/kubernetes/pki/etcd/peer.key --write-out table
+-----------------------------+------------------+---------+---------+-----------+-----------+------------+
| ENDPOINT | ID | VERSION | DB SIZE | IS LEADER | RAFT TERM | RAFT INDEX |
+-----------------------------+------------------+---------+---------+-----------+-----------+------------+
| https://192.168.12.108:2379 | 2db31a5d67ec1034 | 3.5.1 | 7.4 MB | false | 4 | 72291872 |
| https://192.168.12.107:2379 | 471323846709334f | 3.5.1 | 9.0 MB | false | 4 | 72291872 |
| https://192.168.12.109:2379 | 42efe7cca897d765 | 3.5.1 | 9.2 MB | true | 4 | 72291872 |
+-----------------------------+------------------+---------+---------+-----------+-----------+------------+
由于从 kube-apiserver 的日志中同样无法提取出能够帮助解决问题的有用信息,起初我们只能猜测可能是 kube-apiserver 的缓存更新异常导致的。正当我们要从这个切入点去解决问题时,该同事反馈了一个更诡异的问题——自己新创建的 Pod,通过 kubectl查询 Pod 列表,突然消失了!纳尼?这是什么骚操作?经过我们多次测试查询发现,通过 kubectl 来 list pod 列表,该 pod 有时候能查到,有时候查不到。那么问题来了,K8s api 的 list 操作是没有缓存的,数据是 kube-apiserver 直接从 etcd 拉取返回给客户端的,难道是 etcd 本身出了问题?
众所周知,etcd 本身是一个强一致性的 KV 存储,在写操作成功的情况下,两次读请求不应该读取到不一样的数据。怀着不信邪的态度,我们通过 etcdctl 直接查询了 etcd 集群状态和集群数据,返回结果显示 3 个节点状态都正常,且 RaftIndex 一致,观察 etcd 的日志也并未发现报错信息,唯一可疑的地方是 3 个节点的 dbsize 差别较大。接着,我们又将 client 访问的 endpoint 指定为不同节点地址来查询每个节点的 key 的数量,结果发现 3 个节点返回的 key 的数量不一致,甚至两个不同节点上 Key 的数量差最大可达到几千!而直接通过 etcdctl 查询刚才创建的 Pod,发现访问某些 endpoint 能够查询到该 pod,而访问其他 endpoint 则查不到。
至此,基本可以确定 etcd 集群的节点之间确实存在数据不一致现象。
解决思路
- 备份正常节点的etcd数据和对应的数据目录
- 停止异常数据etcd
- 正常etcd节点,删除异常member
- 清除member/ wal/目录下的数据
- 异常节点重新加入集群
- 启动etcd服务就可以了
解决实践: 此处将weiyigeek-109中的etcd数据备份恢复到weiyigeek-107与weiyigeek-108节点中。
# 1.导出 DB SIZE 最大的 etcd 节点数据(操作前一定要备份各 etcd 节点数据)。
etcdctl --endpoints=https://192.168.12.107:2379 snapshot save /backup/etcd-107-snapshot.db
etcdctl --endpoints=https://192.168.12.108:2379 snapshot save /backup/etcd-108-snapshot.db
etcdctl --endpoints=https://192.168.12.109:2379 snapshot save /backup/etcd-109-snapshot.db # 最完整的etd数据节点(它是leader)
# 补充:如果需要迁移leader,执行如下命令将leader交于weiyigeek-109
etcdctl --cacert=/etc/kubernetes/pki/etcd/ca.crt --cert=/etc/kubernetes/pki/etcd/peer.crt --key=/etc/kubernetes/pki/etcd/peer.key --endpoints=192.168.12.107:2379 move-leader 42efe7cca897d765
# 2.备份异常节点上 etcd 目录
mv /var/lib/etcd /var/lib/etcd.bak
# 或者压缩备份
tar -czvf etcd-20230308bak.taz.gz /var/lib/etcd
# 3.从etcd集群中移除异常节点,注意此处指定的endpoints是正常的leader节点。
$ etcdctl --cacert=/etc/kubernetes/pki/etcd/ca.crt --cert=/etc/kubernetes/pki/etcd/peer.crt --key=/etc/kubernetes/pki/etcd/peer.key --endpoints=192.168.12.109:2379 member list
# 异常 异常的 weiyigeek-107 节点
$ etcdctl --cacert=/etc/kubernetes/pki/etcd/ca.crt --cert=/etc/kubernetes/pki/etcd/peer.crt --key=/etc/kubernetes/pki/etcd/peer.key --endpoints=192.168.12.109:2379 member remove 2db31a5d67ec1034
# 4.修改weiyigeek-107节点上的etcd服务, 先将该节点节点再次加入etcd集群。。
$ mv /etc/kubernetes/manifests/etcd.yaml /etc/kubernetes/manifests-backup/etcd.yaml
$ vim /etc/kubernetes/manifests-backup/etcd.yaml
- --initial-cluster=weiyigeek-109=https://192.168.12.109:2380,weiyigeek-108=https://192.168.12.108:2380,weiyigeek-107=https://192.168.12.107:2380
- --initial-cluster-state=existing
$ kubectl exec -it -n kube-system etcd-weiyigeek-107 sh
$ export ETCDCTL_API=3
$ ETCD_ENDPOINTS=192.168.12.109:2379
$ etcdctl --cacert=/etc/kubernetes/pki/etcd/ca.crt --cert=/etc/kubernetes/pki/etcd/peer.crt --key=/etc/kubernetes/pki/etcd/peer.key --endpoints=$ETCD_ENDPOINTS member add weiyigeek-107 --peer-urls=https://192.168.12.107:2380
Member 8cecf1f50b91e502 added to cluster 71d4ff56f6fd6e70
# 5.加入到集群后,启动 weiyigeek-107 节点中的 etcd 服务
mv /etc/kubernetes/manifests-backup/etcd.yaml /etc/kubernetes/manifests/etcd.yaml
验证结果
解决之后可以发现节点启动正常,数据也变成一致,各Etcd节点keys的数量是一致的都是1976,且数据大小都变成一致了。
$ for count in {107..109};do ETCDCTL_API=3 etcdctl get / --prefix --keys-only --endpoints=https://192.168.12.${count}:2379 --cacert=/etc/kubernetes/pki/etcd/ca.crt --cert=/etc/kubernetes/pki/etcd/peer.crt --key=/etc/kubernetes/pki/etcd/peer.key | wc -l;done
1976
$ for count in {107..109};do ETCDCTL_API=3 etcdctl endpoint status --endpoints=https://192.168.12.${count}:2379 --cacert=/etc/kubernetes/pki/etcd/ca.crt --cert=/etc/kubernetes/pki/etcd/peer.crt --key=/etc/kubernetes/pki/etcd/peer.key ;done
https://192.168.12.107:2379, 8cecf1f50b91e502, 3.5.1, 9.2 MB, false, 9, 72291872
https://192.168.12.108:2379, 35672f075108e36a, 3.5.1, 9.2 MB, false, 9, 72291872
https://192.168.12.109:2379, 42efe7cca897d765, 3.5.1, 9.2 MB, true, 9, 72291872
至此,K8S集群中etcd数据不一致解决与恢复完毕,希望大家多多支持!
原文地址: https://blog.weiyigeek.top/2022/10-15-690.html
本文至此完毕,更多技术文章,尽情期待下一章节!
温馨提示:唯一极客技术博客文章在线浏览【极客全栈修炼】小程序上线了,涉及网络安全、系统运维、应用开发、物联网实战、全栈文章,希望和大家一起学习进步,欢迎浏览交流!(希望大家多多提提意见)
专栏书写不易,如果您觉得这个专栏还不错的,请给这篇专栏 【点个赞、投个币、收个藏、关个注,转个发,留个言】(人间六大情),这将对我的肯定,谢谢!。
温馨提示: 由于作者水平有限,本章错漏缺点在所难免,希望读者批评指正,并请在文章末尾留下您宝贵的经验知识,联系邮箱地址 master@weiyigeek.top 或者关注公众号 WeiyiGeek 联系我。
容灾恢复 | 记一次K8S集群中etcd数据快照的备份恢复实践的更多相关文章
- 万级K8s集群背后etcd稳定性及性能优化实践
背景与挑战 随着腾讯自研上云及公有云用户的迅速增长,一方面,腾讯云容器服务TKE服务数量和核数大幅增长, 另一方面我们提供的容器服务类型(TKE托管及独立集群.EKS弹性集群.edge边缘计算集群.m ...
- 万级K8s集群背后 etcd 稳定性及性能优化实践
1背景与挑战随着腾讯自研上云及公有云用户的迅速增长,一方面,腾讯云容器服务TKE服务数量和核数大幅增长, 另一方面我们提供的容器服务类型(TKE托管及独立集群.EKS弹性集群.edge边缘计算集群.m ...
- k8s 集群中的etcd故障解决
一次在k8s集群中创建实例发现etcd集群状态出现连接失败状况,导致创建实例失败.于是排查了一下原因. 问题来源 下面是etcd集群健康状态: [root@docker01 ~]# cd /opt/k ...
- k8s集群中遇到etcd集群故障的排查思路
一次在k8s集群中创建实例发现etcd集群状态出现连接失败状况,导致创建实例失败.于是排查了一下原因. 问题来源 下面是etcd集群健康状态: 1 2 3 4 5 6 7 8 9 10 11 [roo ...
- 在 Nebula K8s 集群中使用 nebula-spark-connector 和 nebula-algorithm
本文首发于 Nebula Graph Community 公众号 解决思路 解决 K8s 部署 Nebula Graph 集群后连接不上集群问题最方便的方法是将 nebula-algorithm / ...
- 【K8S学习笔记】Part2:获取K8S集群中运行的所有容器镜像
本文将介绍如何使用kubectl列举K8S集群中运行的Pod内的容器镜像. 注意:本文针对K8S的版本号为v1.9,其他版本可能会有少许不同. 0x00 准备工作 需要有一个K8S集群,并且配置好了k ...
- 假如Kafka集群中一个broker宕机无法恢复,应该如何处理?
假如Kafka集群中一个broker宕机无法恢复, 应该如何处理? 今天面试时遇到这个问题, 网上资料说添加新的broker, 是不会自动同步旧数据的. 笨办法 环境介绍 三个broker的集群, z ...
- 将 master 节点服务器从 k8s 集群中移除并重新加入
背景 1 台 master 加入集群后发现忘了修改主机名,而在 k8s 集群中修改节点主机名非常麻烦,不如将 master 退出集群改名并重新加入集群(前提是用的是高可用集群). 操作步骤 ssh 登 ...
- k8s集群中部署prometheus server
1.概述 本文档主要介绍如何在k8s集群中部署prometheus server用来作为监控的数据采集服务器,这样做可以很方便的对k8s集群中的指标.pod的.节点的指标进行采集和监控. 2.下载镜像 ...
- 【DB宝46】NoSQL数据库之CouchBase简介、集群搭建、XDCR同步及备份恢复
目录 一. CouchBase概述 1.1.简述 1.2.CouchDB和CouchBase比对 1.2.1.CouchDB和CouchBase的相同之处 1.2.2.CouchDB和CouchBas ...
随机推荐
- MySQL 常用命令(4)------mysqladmin命令详解
mysqladmin是一个执行管理操作的客户端程序.它可以用来检查服务器的配置和当前状态.创建和删除数据库等. mysqladmin 工具的使用格式: mysqladmin [option] comm ...
- ubuntu 20.04 更新镜像源
首先备份系统自带的源,以防修改错误! 命令: sudo cp /etc/apt/sources.list /etc/apt/sources.list.back 1.修改系统镜像源 文件目录:/etc/ ...
- 夜神模拟器连接不上adb的解决办法
1.夜神模拟器连接不上adb的解决办法 转自 (https://www.jianshu.com/p/6041e64518a8) 最近给模拟器升级了版本,用了一段时间后,突然发现通过adb device ...
- Coursera Programming Languages, Part C 华盛顿大学 Week 3
整个系列课程的最后一小结! 介绍了之前在 interface 中所提到的 subtype 系统以及其与 ML 中 generics 的不同 introduction to subtyping 在之前的 ...
- Kubernetes--管理Pod对象的容器(2)
自定义运行的容器化应用 由Docker镜像启动容器时运行的应用程序在相应的Dockerfile中由ENTRYPOINT指令进行定义,传递给程序的参数则通过CMD指令指定,ENTRYPOINT指令不存在 ...
- 「SOL」序列 (LOJ/NOI2019)
准备写新博客的时候发现自己草稿箱里还有一篇咕了十几天的题解 思路挂在了费用流之前-- 题面 > Link LOJ #3158 解析 这道题的本质是一个二分图带权匹配的问题,一个经典的做法是直接做 ...
- Mysql-------查询各科成绩前三名的记录
查询各科成绩前三名的记录 分析如下: 1 select s.* from (select distinct Student.*,a.c_id, a.s_score, count( distinct b ...
- 机器学习之pandas介绍
pandas简介 pandas全称python Data Analysis Library,是基于numpy的一种工具,pandas纳入了大量库和一些标准的数据模型,提供了高效的操作大型数据集所需的工 ...
- Flutter之EdgeInsets
EdgeInsets我们看看EdgeInsets提供的便捷方法: fromLTRB(double left, double top, double right, doublebottom):分别指定四 ...
- 作业三:CART回归树
作业三:CART回归树 20大数据三班 博客链接 学号 201613336 问题一: 表1为拖欠贷款人员训练样本数据集,使用CART算法基于该表数据构造决策树模型,并使用表2中测试样本集确定剪枝后的最 ...