共计 1584 个字符，预计需要花费 4 分钟才能阅读完成。
背景
因为 k8s 集群的资源有余，将局部服务器关机扩容再启动，发现局部 pod 始终没有起来。
排查过程
查看 pod 的日志，发现解析集群内的域名失败，揣测是 dns 的问题。
查看 dns 相干的 pod，发现 nodelocaldns 始终在重启。
root@master1:~# k get pods --all-namespaces | grep dns
kube-system coredns-74d59cc5c6-9brnf 1/1 Running 0 15d
kube-system coredns-74d59cc5c6-g46rf 1/1 Running 0 15d
kube-system nodelocaldns-bwnml 1/1 Running 0 15d
kube-system nodelocaldns-f8tmj 0/1 CrashLoopBackOff 2926 12d
kube-system nodelocaldns-rtngg 0/1 CrashLoopBackOff 44 15d 登录对应的 node，netstat -ntple | grep 53 查看 53 端口的占用状况，发现被 named 过程占用。
将 named 过程 kill 掉，在 /lib/systemd/system 目录下定位到是 bind9 服务开机自启动，并将其禁用掉，实现修复工作。
root@node1:/lib/systemd/system# grep -rn named ./
./bind9-pkcs11.service:3:Documentation=man:named(8)
./bind9-pkcs11.service:8:Environment=KRB5_KTNAME=/etc/bind/named.keytab
./bind9-pkcs11.service:10:ExecStart=/usr/sbin/named-pkcs11 -f -u bind
./bind9-resolvconf.service:3:Documentation=man:named(8) man:resolvconf(8)
./bind9-resolvconf.service:11:ExecStart=/bin/sh -c 'echo nameserver 127.0.0.1 | /sbin/resolvconf -a lo.named'
./bind9-resolvconf.service:12:ExecStop=/sbin/resolvconf -d lo.named
./systemd-hostnamed.service:12:Documentation=man:systemd-hostnamed.service(8) man:hostname(5) man:machine-info(5)
./systemd-hostnamed.service:13:Documentation=https://www.freedesktop.org/wiki/Software/systemd/hostnamed
./systemd-hostnamed.service:16:ExecStart=/lib/systemd/systemd-hostnamed
./bind9.service:3:Documentation=man:named(8)
./bind9.service:10:ExecStart=/usr/sbin/named -f $OPTIONS
root@node1:/lib/systemd/system# systemctl disable bind9
Synchronizing state of bind9.service with SysV service script with /lib/systemd/systemd-sysv-install.
Executing: /lib/systemd/systemd-sysv-install disable bind9正文完