How to solve the error when starting docker

Solution to the error when starting docker: 1. Open docker and add the content "OPTIONS="--selinux-enabled...""; 2. Clear the filter table of iptables; 3. Execute "docker -storage-setup" command and so on.

The operating environment of this article: CentOS 7.2 system, Docker version 18.04.0, Dell G3 computer.

How to solve the error when starting docker?

Summary of error reports when Docker starts up

Eight common Docker faults

Error report one: error initializing graphdriver

Docker start-up error reports

The system is CentOS 7.2
The system kernel and docker version are as follows:

[root@docker ~]# uname -r
3.10.0-327.el7.x86_64
[root@docker ~]# 
[root@docker ~]# 
[root@docker ~]# 
[root@docker ~]# docker version
Client:
 Version:   18.04.0-ce
 API version:   1.37
 Go version:    go1.9.4
 Git commit:    3d479c0
 Built: Tue Apr 10 18:21:36 2018
 OS/Arch:   linux/amd64
 Experimental:  false
 Orchestrator:  swarm

Server:
 Engine:
  Version:  18.04.0-ce
  API version:  1.37 (minimum version 1.12)
  Go version:   go1.9.4
  Git commit:   3d479c0
  Built:    Tue Apr 10 18:25:25 2018
  OS/Arch:  linux/amd64
  Experimental: false

The startup error message is as follows:

[root@docker ~]# systemctl start docker
Job for docker.service failed because the control process exited with error code. See "systemctl status docker.service" and "journ
[root@docker ~]# 
[root@docker ~]# 
[root@docker ~]# 
[root@docker ~]# systemctl status docker
● docker.service - Docker Application Container Engine
   Loaded: loaded (/usr/lib/systemd/system/docker.service; disabled; vendor preset: disabled)
   Active: failed (Result: start-limit) since 日 2018-04-22 20:52:39 CST; 5s ago
     Docs: https://docs.docker.com
  Process: 4810 ExecStart=/usr/bin/dockerd (code=exited, status=1/FAILURE)
 Main PID: 4810 (code=exited, status=1/FAILURE)

4月 22 20:52:39 docker.cgy.com systemd[1]: Failed to start Docker Application Container Engine.
4月 22 20:52:39 docker.cgy.com systemd[1]: Unit docker.service entered failed state.
4月 22 20:52:39 docker.cgy.com systemd[1]: docker.service failed.
4月 22 20:52:39 docker.cgy.com systemd[1]: docker.service holdoff time over, scheduling restart.
4月 22 20:52:39 docker.cgy.com systemd[1]: start request repeated too quickly for docker.service
4月 22 20:52:39 docker.cgy.com systemd[1]: Failed to start Docker Application Container Engine.
4月 22 20:52:39 docker.cgy.com systemd[1]: Unit docker.service entered failed state.
4月 22 20:52:39 docker.cgy.com systemd[1]: docker.service failed.

The specific cause of the error is not seen from the above error message. . Then I used dockerd to start it directly, and I saw an error message at the bottom of the output information, as follows:

[root@docker ~]# dockerd
INFO[2018-04-22T21:12:46.111704443+08:00] libcontainerd: started new docker-containerd process  pid=5903
INFO[0000] starting containerd                           module=containerd revision=773c489c9c1b21a6d78b5c538cd395416ec50f88 version=v1.0.3

。。。。。。省略一部分输出。。。。。。

INFO[0000] loading plugin "io.containerd.grpc.v1.introspection"...  module=containerd type=io.containerd.grpc.v1
INFO[0000] serving...                                    address="/var/run/docker/containerd/docker-containerd-debug.sock" module="containerd/debug"
INFO[0000] serving...                                    address="/var/run/docker/containerd/docker-containerd.sock" module="containerd/grpc"
INFO[0000] containerd successfully booted in 0.002763s   module=containerd
Error starting daemon: error initializing graphdriver: overlay: the backing xfs filesystem is formatted without d_type support, which leads to incorrect behavior. Reformat the filesystem with ftype=1 to en d_type support. Backing filesystems without d_type support are not supported.

According to the last error Error starting daemon: Searched this blog and got the solution.
https://blog.csdn.net/liu9718214/article/details/79134900

The specific solution is:

##vim /etc/sysconfig/docker Add the following content:

OPTIONS="--selinux-enabled --log-driver=journald --signature-verification=false"

/etc/docker/daemon.jsonAdd the following content:

{
    "registry-mirrors": ["http://4a1df5ef.m.daocloud.io"],   # 是用来pull容器加速用的，跟此次问题无关。
    "storage-driver": "devicemapper"    # 解决此次问题
}

Then restart docker and solve the problem smoothly:

[root@docker ~]# systemctl restart docker
[root@docker ~]# 
[root@docker ~]# 
[root@docker ~]# ps aux | grep docker
root       5922  1.7  1.6 528432 62568 ?        Ssl  21:15   0:00 /usr/bin/dockerd
root       5927  1.1  0.5 356984 22100 ?        Ssl  21:15   0:00 docker-containerd --config /var/run/docker/containerd/containerd.toml
root       6028  0.0  0.0 112664   964 pts/0    S+   21:15   0:00 grep --color=auto docker

Error 2: iptables failed

FirewallD

firewalld is introduced in CentOS-7. The bottom layer of firewall is to use iptables for data filtering and is built on iptables. This may cause Conflict with Docker.

When firewalld starts or restarts, the DOCKER rules will be removed from iptables, thus affecting the normal work of Docker.

When you are using Systemd, firewalld will start before Docker, but if you start or restart firewalld after Docker starts, you need to restart the Docker process.

System:

[root@controller ~]# cat /etc/redhat-release 
CentOS Linux release 7.0.1406 (Core)

The error message is as follows:

[root@controller ~]# docker run -it -P docker.io/nginx
/usr/bin/docker-current: Error response from daemon: driver failed programming external connectivity on endpoint gloomy_kirch (10289e7a87e65771da90cda531951b7339bee9cb5953474460451cd48013aff0): iptables failed: iptables --wait -t nat -A DOCKER -p tcp -d 0/0 --dport 32810 -j DNAT --to-destination 172.17.0.2:80 ! -i docker0: iptables: No chain/target/match by that name.
 (exit status 1).

This is because the container was successfully started once before running this time. During the last visit, due to the firewall The problem prevented normal access to Nginx, so I cleared the filter table of iptables, restarted iptables, and then ran it again, the above error was reported.

Solution

Restart the firewall

#CentOS 7下执行
[root@controller ~]# systemctl restart firewalld

Restart the docker daemon process

[root@controller ~]# systemctl restart docker

Run an nginx in the container again and there will be no error

[root@controller ~]# docker run -it --name nginx -p 80:80 -v /www:/wwwroot docker.io/nginx /bin/bash
root@a8a92c8f7760:/#

Error three: Unable to take ownership of thin-pool

Docker daemon failed to start: Unable to take ownership of thin-pool

Apr 27 13:51:59 master systemd: Started Docker Storage Setup.
Apr 27 13:51:59 master systemd: Starting Docker Application Container Engine...
Apr 27 13:51:59 master dockerd-current: time="2018-04-27T13:51:59.088441356+08:00" level=warning msg="could not change group /var/run/docker.sock to docker: group docker not found"
Apr 27 13:51:59 master dockerd-current: time="2018-04-27T13:51:59.091166189+08:00" level=info msg="libcontainerd: new containerd process, pid: 20930"
Apr 27 13:52:00 master dockerd-current: Error starting daemon: error initializing graphdriver: devmapper: Unable to take ownership of thin-pool (docker--vg-docker--pool) that already has used data blocks
Apr 27 13:52:00 master systemd: docker.service: main process exited, code=exited, status=1/FAILURE
Apr 27 13:52:00 master systemd: Failed to start Docker Application Container Engine.
Apr 27 13:52:00 master systemd: Unit docker.service entered failed state.
Apr 27 13:52:00 master systemd: docker.service failed

Reason: /var/lib/ Metadata in docker/devicemapper/metadata/ is lost

workaround：

https://bugzilla.redhat.com/show_bug.cgi?id=1321640#c5

Eric Paris 2016-04-27 08:20:10 EDT

I feel like the kcs kinda misses telling users the actual problem. Nor does it really make it clear the solution.

IF you are using device mapper (instead of loopback) /var/lib/docker contains metadata informing docker about the contents of the device mapper storage area. If you delete /var/lib/docker that metadata is lost. Docker is then able to detect that the thin pool has data but docker is unable to make use of that information. The only solution is to delete the thin pool and recreate it so that both the thin pool and the metadata in /var/lib/docker will be empty.

Solution:

• Execute command:

  rm -rf /var/lib/docker/*

• Execute the command:

  rm -rf /etc/sysconfig/docker-storage

• Execute the command:

  lvremove /dev/docker-vg/docker-pool

• Use existing docker-vg LVM volume group:

cat  /etc/sysconfig/docker-storage-setup
VG=docker-vg
EOF

• Execute the command:

  docker-storage-setup

##Restart docker:
• systemctl start docker

Error four: write /sys/fs/cgroup/cpuset/docker/cpuset.cpus: invalid argument

The following error is reported when docker run runs the container:

[root@backup-system cpu]# docker run -ti --name hkp_ubuntu  --cpuset-cpus=0-3 ubuntu bash
docker: Error response from daemon: OCI runtime create failed: container_linux.go:370: starting container process caused: process_linux.go:326: applying cgroup configuration for process caused: failed to write "0-3\n" to "/sys/fs/cgroup/cpuset/docker/cpuset.cpus": write /sys/fs/cgroup/cpuset/docker/cpuset.cpus: invalid argument: unknown.

This error is because the cpu of this cgroup is being used by other cgroups, so exclusive cannot be set.

Therefore, you need to check and adjust the cpuset.cpus of each cgroup first to ensure that the cpu used by the current cgroup is indeed only allocated to it. Then you can set cpu_exclusive at this time.

The current specific reason is that during the experiment, a new container directory was created in

/sys/fs/cgroup/cpuset/

, and container/cpuset.cpus was set to 0-3<pre class="brush:php;toolbar:false">[root@backup-system docker]# cat /sys/fs/cgroup/cpuset/container/cpuset.cpus  0-3</pre>

Solution

:After setting /sys/fs/cgroup/cpuset/container/cpuset.cpus
to empty, the above problem is solved. Recommended learning: "

docker video tutorial

"

The above is the detailed content of How to solve the error when starting docker. For more information, please follow other related articles on the PHP Chinese website!