部署TLS认证的etcd集群¶
分布式高可用的etcd集群,在生产环境上需要启用TLS认证来加强安全性。在完成 部署etcd集群 之后,我在部署 K3s - 轻量级Kubernetes 中,采用 K3s高可用etcd 实现。
实践环境¶
在 树莓派堆叠 环境采用3台 树莓派Raspberry Pi 3 硬件部署3节点 etcd - 分布式kv存储 集群:
主机IP |
主机名 |
---|---|
192.168.7.11 |
x-k3s-m-1 |
192.168.7.12 |
x-k3s-m-2 |
192.168.7.13 |
x-k3s-m-3 |
TLS证书¶
TLS证书采用 cfssl
工具构建,完整步骤见 etcd集群TLS设置 。分别获得:
CA:
ca-key.pem ca.csr ca.pem
服务器证书:
server-key.pem server.csr server.pem
点对点证书:
x-k3s-m-1-key.pem x-k3s-m-1.csr x-k3s-m-1.pem x-k3s-m-2.csr x-k3s-m-2.json x-k3s-m-2.pem x-k3s-m-3.csr x-k3s-m-3.json x-k3s-m-3.pem
客户端证书:
client-key.pem client.csr client.pem
证书分发¶
为方便使用ssh/scp进行管理,首先采用 ssh密钥 的 shell环境解决ssh-agent对会话的要求 结合 ssh多路传输multiplexing加速 ,这样可以不必输入密码就可以ssh/scp到集群服务器
配置好 边缘云计算架构 的 配置DNSmasq对不同子网提供不同域名扩展(expand-hosts) ,提供正确的域名解析,这样后面配置
etcd
可以正确获得主机名解析使用 etcd集群TLS设置 方法完成上述证书生成后,使用以下脚本进行分发:
cat << EOF > etcd_hosts
x-k3s-m-1
x-k3s-m-2
x-k3s-m-3
EOF
cat << EOF > prepare_etcd.sh
if [ -d /tmp/etcd_tls ];then
rm -rf /tmp/etcd_tls
mkdir /tmp/etcd_tls
else
mkdir /tmp/etcd_tls
fi
if [ ! -d /etc/etcd/ ];then
sudo mkdir /etc/etcd
fi
EOF
for host in `cat etcd_hosts`;do
scp prepare_etcd.sh $host:/tmp/
ssh $host 'sh /tmp/prepare_etcd.sh'
done
for host in `cat etcd_hosts`;do
scp ${host}.pem ${host}:/tmp/etcd_tls/
scp ${host}-key.pem ${host}:/tmp/etcd_tls/
scp ca.pem ${host}:/tmp/etcd_tls/
scp server.pem ${host}:/tmp/etcd_tls/
scp server-key.pem ${host}:/tmp/etcd_tls/
scp client.csr ${host}:/tmp/etcd_tls/
scp client.pem ${host}:/tmp/etcd_tls/
scp client-key.pem ${host}:/tmp/etcd_tls/
ssh $host 'sudo cp /tmp/etcd_tls/* /etc/etcd/;sudo chown etcd:etcd /etc/etcd/*'
done
执行脚本:
sh deploy_etcd_certificates.sh
这样在 etcd
主机上分别有对应主机的配置文件 /etc/etcd
目录下有(以下案例是 x-k3s-m-1
):
ca.pem
server-key.pem
server.pem
x-k3s-m-1-key.pem
x-k3s-m-1.pem
配置启动服务脚本¶
Systemd进程管理器 启动etcd脚本¶
… 待完善
OpenRC 启动etcd脚本¶
备注
参考alpine linux软件仓库 etcd
和 etcd-openrc
软件包配置方法,我在 OpenRC 配置etcd时,采用 /etc/etcd/conf.yml
配置文件来配置etcd。
需要注意 etcd配置规则 : 如果提供配置文件,则运行参数和环境变量都不会生效!!!
当然,也可以采用命令行参数(见上文 Systemd进程管理器 启动etcd脚本 ),或者采用环境变量。
从惯例来看,大多数网上配置都采用命令行参数,部分采用环境变量。不过,我觉得对于固定配置,特别是不依赖 Systemd进程管理器 这样的启动器,使用配置文件是最通用的。
Alpine Linux 的 edge/testing仓库提供etcd-openrc ,包含了 openrc 启动脚本以及配置:
/etc/init.d/etcd
/etc/conf.d/etcd
此外 edge/testing仓库etcd 包含了 etcd 配置文件 /etc/etcd/conf.yml
所以,在测试主机 x-dev
上 apk添加软件包 ( edge/testing
)
sudo apk add etcd-openrc etcd --update-cache --repository http://dl-cdn.alpinelinux.org/alpine/edge/testing/ --allow-untrusted
可以在此基础上做自定义配置,
准备配置文件
conf.yml
(这个配置文件是 edge/testing仓库etcd 的etcd 配置文件/etc/etcd/conf.yml
基础上修订,增加了配置占位符方便后续通过脚本修订):
# This is the configuration file for the etcd server.
# Human-readable name for this member.
name: 'NODENAME'
# Path to the data directory.
data-dir: /var/lib/etcd
# Path to the dedicated wal directory.
wal-dir:
# Number of committed transactions to trigger a snapshot to disk.
snapshot-count: 10000
# Time (in milliseconds) of a heartbeat interval.
heartbeat-interval: 100
# Time (in milliseconds) for an election to timeout.
election-timeout: 1000
# Raise alarms when backend size exceeds the given quota. 0 means use the
# default quota.
quota-backend-bytes: 0
# List of comma separated URLs to listen on for peer traffic.
listen-peer-urls: https://NODEIP:2380
# List of comma separated URLs to listen on for client traffic.
listen-client-urls: https://NODEIP:2379,https://127.0.0.1:2379
# Maximum number of snapshot files to retain (0 is unlimited).
max-snapshots: 5
# Maximum number of wal files to retain (0 is unlimited).
max-wals: 5
# Comma-separated white list of origins for CORS (cross-origin resource sharing).
cors:
# List of this member's peer URLs to advertise to the rest of the cluster.
# The URLs needed to be a comma-separated list.
initial-advertise-peer-urls: https://NODENAME.DOMAIN:2380
# List of this member's client URLs to advertise to the public.
# The URLs needed to be a comma-separated list.
advertise-client-urls: https://NODENAME.DOMAIN:2379
# Discovery URL used to bootstrap the cluster.
discovery:
# Valid values include 'exit', 'proxy'
discovery-fallback: 'proxy'
# HTTP proxy to use for traffic to discovery service.
discovery-proxy:
# DNS domain used to bootstrap initial cluster.
discovery-srv:
# Initial cluster configuration for bootstrapping.
initial-cluster: NODE1=https://NODE1.DOMAIN:2380,NODE2=https://NODE2.DOMAIN:2380,NODE3=https://NODE3.DOMAIN:2380
# Initial cluster token for the etcd cluster during bootstrap.
initial-cluster-token: 'INIT-TOKEN'
# Initial cluster state ('new' or 'existing').
initial-cluster-state: 'new'
# Reject reconfiguration requests that would cause quorum loss.
strict-reconfig-check: false
# Accept etcd V2 client requests
enable-v2: true
# Enable runtime profiling data via HTTP server
enable-pprof: true
# Valid values include 'on', 'readonly', 'off'
proxy: 'off'
# Time (in milliseconds) an endpoint will be held in a failed state.
proxy-failure-wait: 5000
# Time (in milliseconds) of the endpoints refresh interval.
proxy-refresh-interval: 30000
# Time (in milliseconds) for a dial to timeout.
proxy-dial-timeout: 1000
# Time (in milliseconds) for a write to timeout.
proxy-write-timeout: 5000
# Time (in milliseconds) for a read to timeout.
proxy-read-timeout: 0
client-transport-security:
# Path to the client server TLS cert file.
cert-file: /etc/etcd/server.pem
# Path to the client server TLS key file.
key-file: /etc/etcd/server-key.pem
# Enable client cert authentication.
client-cert-auth: true
# Path to the client server TLS trusted CA cert file.
trusted-ca-file: /etc/etcd/ca.pem
# Client TLS using generated certificates
auto-tls: true
peer-transport-security:
# Path to the peer server TLS cert file.
cert-file: /etc/etcd/NODENAME.pem
# Path to the peer server TLS key file.
key-file: /etc/etcd/NODENAME-key.pem
# Enable peer client cert authentication.
client-cert-auth: true
# Path to the peer server TLS trusted CA cert file.
trusted-ca-file: /etc/etcd/ca.pem
# Peer TLS using generated certificates.
auto-tls: true
# Enable debug-level logging for etcd.
debug: false
logger: zap
# Specify 'stdout' or 'stderr' to skip journald logging even when running under systemd.
log-outputs: [stderr]
# Force to create a new one member cluster.
force-new-cluster: false
auto-compaction-mode: periodic
auto-compaction-retention: "1"
准备一个为每个管控服务器修订etcd配置的脚本
config_etcd.sh
:
NODENAME=`hostname -s`
NODEIP=`ip addr show eth0 | grep "inet\b" | awk '{print $2}' | cut -d/ -f1`
NODE1="x-k3s-m-1"
NODE2="x-k3s-m-2"
NODE3="x-k3s-m-3"
DOMAIN="edge.huatai.me"
INITTOKEN="x-k3s"
cd /tmp/etcd_config
sed -i "s/NODENAME/$NODENAME/g" conf.yml
sed -i "s/NODEIP/$NODEIP/g" conf.yml
sed -i "s/INITTOKEN/$INITTOKEN/g" conf.yml
sed -i "s/NODE1/$NODE1/g" conf.yml
sed -i "s/NODE2/$NODE2/g" conf.yml
sed -i "s/NODE3/$NODE3/g" conf.yml
sed -i "s/DOMAIN/$DOMAIN/g" conf.yml
备注
配置文件 conf.yml
中,初始化etcd绑定的url必须使用主机的IP地址,不能使用域名:
# List of comma separated URLs to listen on for peer traffic.
listen-peer-urls: https://192.168.7.11:2380
# List of comma separated URLs to listen on for client traffic.
listen-client-urls: https://192.168.7.11:2379,https://127.0.0.1:2379
如果使用域名,如 x-k3s-m-1.edge.huatai.me
(即使域名解析正确),启动etcd还是会报错:
{"level":"warn","ts":1649611005.237307,"caller":"etcdmain/etcd.go:74","msg":"failed to verify flags","error":"expected IP in URL for binding (http://x-k3s-m-1.edge.huatai.me:2380)"}
执行以下
deploy_etcd_config.sh
:
cat << EOF > etcd_hosts
x-k3s-m-1
x-k3s-m-2
x-k3s-m-3
EOF
cat << EOF > prepare_etcd_config.sh
if [ -d /tmp/etcd_config ];then
rm -rf /tmp/etcd_config
mkdir /tmp/etcd_config
else
mkdir /tmp/etcd_config
fi
if [ ! -d /etc/etcd/ ];then
sudo mkdir /etc/etcd
fi
EOF
for host in `cat etcd_hosts`;do
scp prepare_etcd_config.sh $host:/tmp/
ssh $host 'sh /tmp/prepare_etcd_config.sh'
done
for host in `cat etcd_hosts`;do
scp config_etcd.sh $host:/tmp/etcd_config/
scp conf.yml $host:/tmp/etcd_config/
ssh $host 'sh /tmp/etcd_config/config_etcd.sh'
ssh $host 'sudo cp /tmp/etcd_config/conf.yml /etc/etcd/'
done
sh deploy_etcd_config.sh
然后验证每台管控服务器上 /etc/etcd/config.yml
配置文件中的占位符是否已经正确替换成主机名。正确情况下, /etc/etcd/conf.yml
中对应 占位符
都会被替换成对应主机的IP地址或者域名,例如:
# Human-readable name for this member.
name: 'x-k3s-m-1'
...
# List of comma separated URLs to listen on for peer traffic.
listen-peer-urls: https://192.168.7.11:2380
# List of comma separated URLs to listen on for client traffic.
listen-client-urls: https://192.168.7.11:2379,https://127.0.0.1:2379
...
# Initial cluster configuration for bootstrapping.
initial-cluster: x-k3s-m-1=https://x-k3s-m-1.edge.huatai.me:2380,x-k3s-m-2=https://x-k3s-m-2.edge.huatai.me:2380,x-k3s-m-3=https://x-k3s-m-3.edge.huatai.me:2380
...
准备配置文件
conf.d-etcd
和init.d-etcd
(从alpine linux软件仓库etcd-openrc
软件包提取)
LOGPATH=/var/log/${RC_SVCNAME}
ETCD_CONFIG_FILE=/etc/etcd/conf.yml
#!/sbin/openrc-run
# Copyright 2016 Alpine Linux
# Distributed under the terms of the GNU General Public License v2
# $Id$
supervisor=supervise-daemon
name="$RC_SVCNAME"
description="Highly-available key-value store"
ETCD_DATA_DIR=$(sed -nr 's/^data-dir:\s*(\/.*)/\1/p' $ETCD_CONFIG_FILE)
command=/usr/bin/etcd
command_args="--config-file=${ETCD_CONFIG_FILE}"
: ${output_log:=$LOGPATH/$RC_SVCNAME.log}
: ${error_log:=$LOGPATH/$RC_SVCNAME.log}
command_user="etcd:etcd"
supervise_daemon_args="--chdir $ETCD_DATA_DIR"
depend() {
need net
}
start_pre() {
checkpath -d -m 0775 -o "$command_user" "$LOGPATH"
checkpath -d -m 0700 -o "$command_user" "$ETCD_DATA_DIR"
}
然后执行以下
deploy_etcd_service.sh
:
cat << EOF > etcd_hosts
x-k3s-m-1
x-k3s-m-2
x-k3s-m-3
EOF
cat << EOF > prepare_etcd_service.sh
if [ -d /tmp/etcd_service ];then
rm -rf /tmp/etcd_service
mkdir /tmp/etcd_service
else
mkdir /tmp/etcd_service
fi
EOF
for host in `cat etcd_hosts`;do
scp prepare_etcd_service.sh $host:/tmp/
ssh $host 'sh /tmp/prepare_etcd_service.sh'
done
for host in `cat etcd_hosts`;do
scp conf.d-etcd $host:/tmp/etcd_service/
scp init.d-etcd $host:/tmp/etcd_service/
ssh $host 'sudo cp /tmp/etcd_service/conf.d-etcd /etc/conf.d/etcd'
ssh $host 'sudo cp /tmp/etcd_service/init.d-etcd /etc/init.d/etcd'
ssh $host 'sudo addgroup -g 1001 etcd && sudo adduser -u 1001 -G etcd -h /dev/null -s /sbin/nologin -D etcd'
done
sh deploy_etcd_service.sh
启动服务:
sudo service etcd start
检查¶
启动
etcd
之后,检查服务进程:ps aux | grep etcd
可以看到:
5665 root 0:00 supervise-daemon etcd --start --stdout /var/log/etcd/etcd.log --stderr /var/log/etcd/etcd.log --user etcd etcd --chdir /var/lib/etcd /usr/bin/etcd -- --config-file=/etc/etcd/conf.yml
5666 etcd 0:01 /usr/bin/etcd --config-file=/etc/etcd/conf.yml
检查日志:
tail -f /var/log/etcd/etcd.log
证书错误排查¶
我在启动 etcd
服务之后,检查 etcd.log
发现有如下报错:
WARNING: 2022/04/11 23:59:00 [core] grpc: addrConn.createTransport failed to connect to {127.0.0.1:2379 127.0.0.1:2379 <nil> 0 <nil>}. Err: connection error: desc = "transport: authentication handshake failed: remote error: tls: bad certificate". Reconnecting...
{"level":"warn","ts":"2022-04-11T23:59:13.567+0800","caller":"embed/config_logging.go:169","msg":"rejected connection","remote-addr":"192.168.7.11:53050","server-name":"","error":"tls: failed to verify client certificate: x509: certificate specifies an incompatible key usage"}
WARNING: 2022/04/11 23:59:13 [core] grpc: addrConn.createTransport failed to connect to {192.168.7.11:2379 192.168.7.11:2379 <nil> 0 <nil>}. Err: connection error: desc = "transport: authentication handshake failed: remote error: tls: bad certificate". Reconnecting...
{"level":"warn","ts":"2022-04-11T23:59:15.891+0800","caller":"embed/config_logging.go:169","msg":"rejected connection","remote-addr":"127.0.0.1:50242","server-name":"","error":"tls: failed to verify client certificate: x509: certificate specifies an incompatible key usage"}
并且,每个服务器上都有类似对应错误
这个问题在 ETCD出现:certificate specifies an incompatible key usage 解决方案 提供了解决思路: 原因参考 3.2/3.3 etcd server with TLS would start with error “tls: bad certificate” #9398 ,是因为在 etcd集群TLS设置 时使用的 ca-config.json
采用了:
...
"server": {
"expiry": "87600h",
"usages": [
"signing",
"key encipherment",
"server auth"
]
},
...
没有添加 "client auth"
,这在早期etcd版本中不是问题,但是在 3.2 版本之后,需要添加,也就是:
...
"server": {
"expiry": "87600h",
"usages": [
"signing",
"key encipherment",
"server auth",
"client auth"
]
},
...
然后重新生成证书
验证etcd集群¶
现在 etcd
集群已经启动,我们使用以下命令检查集群是否正常工作:
curl --cacert ca.pem --cert client.pem --key client-key.pem https://etcd.edge.huatai.me:2379/health
此时返回信息应该是:
{"health":"true","reason":""}
注意,这里访问的域名是 etcd.edge.huatai.me
,原因是 etcd集群TLS设置 配置 server.json
时在证书中指定访问域名是 etcd.edge.huatai.me
(或IP)
{
"CN": "edge k3s etcd",
"hosts": [
"etcd.edge.huatai.me",
"192.168.7.11",
"192.168.7.12",
"192.168.7.13",
"127.0.0.1"
],
"key": {
"algo": "ecdsa",
"size": 256
},
"names": [
{
"C": "CN",
"L": "Shanghai",
"ST": "cloud-atlas"
}
]
}
所以不能使用 x-k3s-m-1.edge.huatai.me
这样的 real server 域名,但是可以使用 192.168.7.11
的IP地址,因为IP地址也配置在证书中
上述 server.json
非常巧妙使用了可以同时解析为多个real server的域名 etcd.edge.huatai.me
,也就是生产环境上,可以配置这个域名轮转到这3台服务器的IP上,或者使用一个 负载均衡 分发到这3个real server上,域名解析绑定到负载均衡的VIP上。
为方便维护,配置
etcdctl
环境变量,添加到用户自己的 profile中:
export ETCDCTL_API=3
export ETCDCTL_ENDPOINTS='https://etcd.edge.huatai.me:2379'
export ETCDCTL_CACERT=/etc/etcd/ca.pem
export ETCDCTL_CERT=/etc/etcd/client.pem
export ETCDCTL_KEY=/etc/etcd/client-key.pem
然后可以检查:
etcdctl member list
输出类似:
9bfd4ef1e72d26, started, x-k3s-m-3, https://x-k3s-m-3.edge.huatai.me:2380, https://x-k3s-m-3.edge.huatai.me:2379, false
7e8d94ba496c072d, started, x-k3s-m-1, https://x-k3s-m-1.edge.huatai.me:2380, https://x-k3s-m-1.edge.huatai.me:2379, false
a01cb65343e64610, started, x-k3s-m-2, https://x-k3s-m-2.edge.huatai.me:2380, https://x-k3s-m-2.edge.huatai.me:2379, false
为方便观察,可以使用表格输出模式:
etcdctl --write-out=table endpoint status
输出显示:
+---------------------------+------------------+---------+---------+-----------+------------+-----------+------------+--------------------+--------+
| ENDPOINT | ID | VERSION | DB SIZE | IS LEADER | IS LEARNER | RAFT TERM | RAFT INDEX | RAFT APPLIED INDEX | ERRORS |
+---------------------------+------------------+---------+---------+-----------+------------+-----------+------------+--------------------+--------+
| https://192.168.7.11:2379 | 7e8d94ba496c072d | 3.5.2 | 20 kB | false | false | 7 | 237 | 237 | |
| https://192.168.7.12:2379 | a01cb65343e64610 | 3.5.2 | 20 kB | false | false | 7 | 237 | 237 | |
| https://192.168.7.13:2379 | 9bfd4ef1e72d26 | 3.5.2 | 20 kB | true | false | 7 | 237 | 237 | |
+---------------------------+------------------+---------+---------+-----------+------------+-----------+------------+--------------------+--------+
检查健康状况:
etcdctl endpoint health
输出显示:
https://192.168.7.13:2379 is healthy: successfully committed proposal: took = 67.98523ms
https://192.168.7.12:2379 is healthy: successfully committed proposal: took = 64.634362ms
https://192.168.7.11:2379 is healthy: successfully committed proposal: took = 67.330493ms
参考¶
Setting up Etcd Cluster with TLS Authentication Enabled 这篇文档非常详细指导了如何使用cfssl工具来生成etcd服务器证书,以及签名客户端证书
How To Setup a etcd Cluster On Linux – Beginners Guide 提供了一个生成 Systemd进程管理器 配置的脚本