由来
客户是深信服的订制系统，基于 centos 改的，排查半天发现居然是文件损坏，而不是 docker 的问题。
环境信息
docker信息:
$ docker info Containers: 0 Running: 0 Paused: 0 Stopped: 0 Images: 2 Server Version: 18.09.3 Storage Driver: overlay2 Backing Filesystem: xfs Supports d_type: true Native Overlay Diff: true Logging Driver: json-file Cgroup Driver: cgroupfs Plugins: Volume: local Network: bridge host macvlan null overlay Log: awslogs fluentd gcplogs gelf journald json-file local logentries splunk syslog Swarm: inactive Runtimes: runc Default Runtime: runc Init Binary: docker-init containerd version: e6b3f5632f50dbc4e9cb6288d911bf4f5e95b18e runc version: 6635b4f0c6af3810594d2770f662f34ddc15b40d init version: fec3683 Security Options: seccomp Profile: default Kernel Version: 3.10.0 Operating System: CentOS Linux 7 (Core) OSType: linux Architecture: x86_64 CPUs: 20 Total Memory: 125.3GiB Name: eds-1f21a854 ID: VZLV:26PU:ZUN6:QQEJ:GW3I:YETT:HMEU:CK6J:SIPL:CHKV:6ASN:5NDF Docker Root Dir: /data/kube/docker Debug Mode (client): false Debug Mode (server): false Registry: https://index.docker.io/v1/ Labels: Experimental: false Insecure Registries: reg.wps.lan:5000 treg.yun.wps.cn 127.0.0.0/8 Registry Mirrors: https://registry.docker-cn.com/ https://docker.mirrors.ustc.edu.cn/ Live Restore Enabled: false Product License: Community Engine
系统信息
$ uname -a Linux eds-1f21a854 3.10.0 #1 SMP Mon Sep 28 12:00:30 CST 2020 x86_64 x86_64 x86_64 GNU/Linux $ cat /etc/os-release NAME="CentOS Linux" VERSION="7 (Core)" ID="centos" ID_LIKE="rhel fedora" VERSION_ID="7" PRETTY_NAME="CentOS Linux 7 (Core)" ANSI_COLOR="0;31" CPE_NAME="cpe:/o:centos:centos:7" HOME_URL="https://www.centos.org/" BUG_REPORT_URL="https://bugs.centos.org/" CENTOS_MANTISBT_PROJECT="CentOS-7" CENTOS_MANTISBT_PROJECT_VERSION="7" REDHAT_SUPPORT_PRODUCT="centos" REDHAT_SUPPORT_PRODUCT_VERSION="7"
服务器信息:
$ cat product_name SUGON-60G16 $ cat sys_vendor SANGFOR $ cat product_version 1.2
排查过程
安装docker服务器挂了
时间2020 10/29 19:51:
实施: 客户这边部署的时候安装docker的时候服务器挂了
我: 开机后/var/log/message有信息吗
实施: 只能恢复快照才能进去，服务器进不了，看不了信息
我: 不恢复快照起不来吗
实施: 是的
到这里我以为是触发了啥内核 bug 直接内核 panic 了服务器起不来。
时间2020 10/30 9:07:
我: 起不来的时候有进控制台去看啥原因起不来吗
实施: 是客户服务器没法查看呢
我: 客户没去看下吗
然后实施直接发来一个向日葵远程连接，我上去后发现不是常规的操作系统，是基于 centos 改过的，没找到/var/log/message，然后手动执行我们的 docker 安装脚本。
bash -x install-docker.sh
然后输出的信息在某一步就没输出了，应该”挂了”，看了下脚本最后一条输出调试信息的后面是启动 docker，应该是启动 docker 触发的。然后很久后还是无法连上和 ping 通，叫实施问问那边现场看看是硬件服务器的话有没有 idrac，ilo 之类的看看 tty 控制台的信息。
现场人员看了下服务器是”正常开机”的，我这边尝试还是连不上，现场问我们的操作是否是改了路由，现场 systemctl 看了下 docker 是起来的。现场那边还是 ping 不通网关。我这边突然想到是不是压根没挂。。。
叫他 uptime -s 看看上次的启动时间，结果压根没重启。。。
然后现场排查到是 iptables 的问题，启动 docker 的时候把他们的规则刷没了。后面他们改了下都放开了。所以前面的启动 docker 把机器挂了实际上是iptables的影响导致网络断开，机器压根没重启。
启动容器挂掉
然后继续，实施说之前同样的其他机器安装 docker 的时候没出现上面的问题，而是启动的时候出现上面的问题，我就手动执行下部署，结果报错。脚本开-x调试看是load 部署镜像的时候报错了。
error during connect: Post http://%2Fvar%2Frun%2Fdocker.sock/v1.39/images/load"htmlcode">$ docker load -i ./kube/images/deploy.tar error during connect: Post http://%2Fvar%2Frun%2Fdocker.sock/v1.39/images/load"htmlcode">systemctl stop docker pkill dockerd dockerd -D另开一个终端执行 load 镜像操作:
$ docker load -i ./kube/images/deploy.tar ab6425526dab: Loading layer [==================================================>] 126.3MB/126.3MB c7fe3ea715ef: Loading layer [==================================================>] 340.3MB/340.3MB 7f7eae7934f7: Loading layer [==================================================>] 3.584kB/3.584kB e99a66706b50: Loading layer [==================================================>] 2.105MB/2.105MB 245b79de6bb7: Loading layer [==================================================>] 283.5MB/283.5MB 2a56a4432cef: Loading layer [==================================================>] 93.18kB/93.18kB 0c2ea71066ab: Loading layer [==================================================>] 276.5kB/276.5kB bb3f6df0f87c: Loading layer [==================================================>] 82.94kB/82.94kB 6f34ead3cef7: Loading layer [==================================================>] 946.7kB/946.7kB c21422dd15f6: Loading layer [==================================================>] 1.97MB/1.97MB 940288517f4c: Loading layer [==================================================>] 458.2kB/458.2kB 0c52f1af61f4: Loading layer [==================================================>] 5.12kB/5.12kB 049e7edd48bc: Loading layer [==================================================>] 1.57MB/1.57MB 73307245d702: Loading layer [==================================================>] 5.632kB/5.632kB f109309d8ffc: Loading layer [==================================================>] 2.175MB/2.175MB Loaded image: xxxxxxxxxxxx.cn/platform/deploy-amd64:ubuntu.16.04 $ docker images REPOSITORY TAG IMAGE ID CREATED SIZE xxxxxxxxxxxx.cn/platform/deploy-amd64 ubuntu.16.04 3ad94a76af5d 3 months ago 734MB调试这边前台日志输出正常
... DEBU[2020-10-30T14:48:39.955963986+08:00] Applied tar sha256:049e7edd48bc46e3dd5edf89c9caa8f0f7efbb41af403c5a54dd4f1008f604a7 to d58edd0d97bb672ef40e82e45c1603ca3ceaad847d9b9fc7c9b0588087019649, size: 1518278 DEBU[2020-10-30T14:48:39.960091040+08:00] Applying tar in /data/kube/docker/overlay2/b044bd592ae800ed071208c6b2f650c5cbdc7452702f56a23b9b4ffe4236ac18/diff storage-driver=overlay2 DEBU[2020-10-30T14:48:40.059510528+08:00] Applied tar sha256:73307245d7021f9627ca0b2cbfeab3aac0b65abfd476f6ec26bb92c75892d7e2 to b044bd592ae800ed071208c6b2f650c5cbdc7452702f56a23b9b4ffe4236ac18, size: 3284 DEBU[2020-10-30T14:48:40.063040538+08:00] Applying tar in /data/kube/docker/overlay2/03918b1d275aa284532b8b9c59ca158409416f904e13cc7084c598ed343e844f/diff storage-driver=overlay2 DEBU[2020-10-30T14:48:40.148209852+08:00] Applied tar sha256:f109309d8ffcb76589ad6389e80335d986b411c80122d990ab00a02a3a916e3e to 03918b1d275aa284532b8b9c59ca158409416f904e13cc7084c598ed343e844f, size: 2072803 ^CINFO[2020-10-30T14:48:55.593523177+08:00] Processing signal 'interrupt' DEBU[2020-10-30T14:48:55.593617229+08:00] daemon configured with a 15 seconds minimum shutdown timeout DEBU[2020-10-30T14:48:55.593638628+08:00] start clean shutdown of all containers with a 15 seconds timeout... DEBU[2020-10-30T14:48:55.594074457+08:00] Unix socket /run/docker/libnetwork/ebd15186e86385c48c4c5508d5f30eb83d5d74e56f09af5c82b6d6d9d63ec8b8.sock doesn't exist. cannot accept client connections DEBU[2020-10-30T14:48:55.594106623+08:00] Cleaning up old mountid : start. INFO[2020-10-30T14:48:55.594157536+08:00] stopping event stream following graceful shutdown error="<nil>" module=libcontainerd namespace=moby DEBU[2020-10-30T14:48:55.594343122+08:00] Cleaning up old mountid : done. DEBU[2020-10-30T14:48:55.594501828+08:00] Clean shutdown succeeded INFO[2020-10-30T14:48:55.594520918+08:00] stopping healthcheck following graceful shutdown module=libcontainerd INFO[2020-10-30T14:48:55.594531978+08:00] stopping event stream following graceful shutdown error="context canceled" module=libcontainerd namespace=plugins.moby DEBU[2020-10-30T14:48:55.594603119+08:00] received signal signal=terminated INFO[2020-10-30T14:48:55.594739890+08:00] pickfirstBalancer: HandleSubConnStateChange: 0xc4201a61b0, TRANSIENT_FAILURE module=grpc INFO[2020-10-30T14:48:55.594751465+08:00] pickfirstBalancer: HandleSubConnStateChange: 0xc4201a61b0, CONNECTING module=grpc看了下systemd的配置没啥特殊的，就很迷，不知道为啥前台运行就能导入，后面实在想不到怎么排查，就怀疑可能是 socket 问题，尝试用 socat 转发成 tcp 试试，结果还是不行(此处应该daemon那加tcp 监听127试试，不应该通过socket，socat最终也是过的socket)
$ socat -d -d TCP-LISTEN:2375,fork,bind=127.0.0.1 UNIX:/var/run/docker.sock 2020/10/30 17:39:58 socat[5201] N listening on AF=2 127.0.0.1:2375 ^[[C2020/10/30 17:42:06 socat[5201] N accepting connection from AF=2 127.0.0.1:35370 on AF=2 127.0.0.1:2375 2020/10/30 17:42:06 socat[5201] N forked off child process 11501 2020/10/30 17:42:06 socat[5201] N listening on AF=2 127.0.0.1:2375 2020/10/30 17:42:06 socat[11501] N opening connection to AF=1 "/var/run/docker.sock" 2020/10/30 17:42:06 socat[11501] N successfully connected from local address AF=1 "\0\0\0\0\0\0 \x0D\x@7\xE9\xEC\x7E\0\0\0\x01\0\0\0\0" 2020/10/30 17:42:06 socat[11501] N starting data transfer loop with FDs [6,6] and [5,5] 2020/10/30 17:42:12 socat[11501] E write(5, 0x55f098349920, 8192): Broken pipe 2020/10/30 17:42:12 socat[11501] N exit(1) 2020/10/30 17:42:12 socat[5201] N childdied(): handling signal 17 $ docker --log-level debug -H tcp://127.0.0.1:2375 load -i kube/images/deploy.tar c7fe3ea715ef: Loading layer [==========================================> ] 286.9MB/340.3MB unexpected EOF最后耗了挺久的，当时忙，去看了下另一个客户的问题，然会回到这边，突发奇想的试试 load 其他镜像，结果可以。。。
$ docker load -i kube/images/pause_3.1.tar e17133b79956: Loading layer [==================================================>] 744.4kB/744.4kB Loaded image: mirrorgooglecontainers/pause-amd64:3.1 $ docker load -i kube/images/tiller_v2.16.1.tar 77cae8ab23bf: Loading layer [==================================================>] 5.815MB/5.815MB 679105aa33fb: Loading layer [==================================================>] 6.184MB/6.184MB 639eab5d05b1: Loading layer [==================================================>] 40.46MB/40.46MB 87e5687e03f2: Loading layer [==================================================>] 41.13MB/41.13MB Loaded image: gcr.io/kubernetes-helm/tiller:v2.16.1 $ docker load -i kube/images/calico_v3.1.3.tar cd7100a72410: Loading layer [==================================================>] 4.403MB/4.403MB ddc4cb8dae60: Loading layer [==================================================>] 7.84MB/7.84MB 77087b8943a2: Loading layer [==================================================>] 249.3kB/249.3kB c7227c83afaf: Loading layer [==================================================>] 4.801MB/4.801MB 2e0e333a66b6: Loading layer [==================================================>] 231.8MB/231.8MB Loaded image: calico/node:v3.1.3 2580685bfb60: Loading layer [==================================================>] 50.84MB/50.84MB Loaded image: calico/kube-controllers:v3.1.3 0314be9edf00: Loading layer [==================================================>] 1.36MB/1.36MB 15db169413e5: Loading layer [==================================================>] 28.05MB/28.05MB 4252efcc5013: Loading layer [==================================================>] 2.818MB/2.818MB 76cf2496cf36: Loading layer [==================================================>] 3.03MB/3.03MB 91d3d3a16862: Loading layer [==================================================>] 2.995MB/2.995MB 18a58488ba3b: Loading layer [==================================================>] 3.474MB/3.474MB 8d8197f49da2: Loading layer [==================================================>] 27.34MB/27.34MB 7520364e0845: Loading layer [==================================================>] 9.216kB/9.216kB b9d064622bd6: Loading layer [==================================================>] 2.56kB/2.56kB Loaded image: calico/cni:v3.1.3只有导入这个的时候才报错
$ docker load -i kube/images/deploy.tar error during connect: Post http://%2Fvar%2Frun%2Fdocker.sock/v1.39/images/load?quiet=0: read unix @->/var/run/docker.sock: read: connection reset by peer然后出包的机器上对比了下这个文件的校验值发现不对。。。。
总结
有个疑问就是为啥前台可以，其次文件损坏导入的时候 docker daemon 居然不刷任何日志直接 connection reset，新版本没测试过这种情况。