systemd "docker.socket" missing on todays update #476
Comments
|
I confirm. Downgrading back manually by downloading 18.06.1 apt remove --purge containerd.io docker-ce docker-ce-cli
wget https://download.docker.com/linux/debian/dists/stretch/pool/stable/amd64/docker-ce_18.06.1~ce~3-0~debian_amd64.deb
dpkg -i docker-ce_18.06.1~ce~3-0~debian_amd64.deb |
|
I confirm also in ubuntu. does the trick. however, i digged into this and : |
|
Also reported in moby/moby#38119 |
|
Same problem for me: Downgrade to |
|
on my rasbian i could not start docker after the update.
i found out that i had a new file in /systemd/system/docker.service.d/override.conf with this content: so wtf i dont know where this file came from, but this is a loop. but even if i remove the last 2 lines i cannot start docker. can also confirm that somewhere in the logs i also read that unix did not get detected, instead it was sth with rtcp or similar, cannot find and remember it :S |
|
Maybe include systemd integration tests before releasing a new version? |
There's tests for this, but unfortunately, the problem with the daemon failing to start won't manifest it self on a fresh install, and only will only be an issue if;
If you're in situation 1.;The unmodified systemd unit file that ships with Docker 18.09.0 can be found here; https://github.com/docker/docker-ce/blob/v18.09.0/components/packaging/systemd/docker.service If you're in situation 2.:If you need socket activation; restore the docker.socket unit-file from the 18.06 release; https://github.com/docker/docker-ce/blob/18.06/components/packaging/deb/systemd/docker.socket If you don't need socket activation; either:
Work is in progress to add back socket activation (see |
i don't have any custom config files and it still fails. maybe also add this test case to your testsuite? install previous version and upgrade. |
|
Trying to reproduce, but it seems to work for me (Ubuntu 16.04 in this case); Version installed before upgrading; docker version
Client:
Version: 18.06.1-ce
API version: 1.38
Go version: go1.10.3
Git commit: e68fc7a
Built: Tue Aug 21 17:24:56 2018
OS/Arch: linux/amd64
Experimental: false
Server:
Engine:
Version: 18.06.1-ce
API version: 1.38 (minimum version 1.12)
Go version: go1.10.3
Git commit: e68fc7a
Built: Tue Aug 21 17:23:21 2018
OS/Arch: linux/amd64
Experimental: falseCheck the service's unit-file for modifications (and/or show overrides/drop-ins); systemctl cat docker.service
# /lib/systemd/system/docker.service
[Unit]
Description=Docker Application Container Engine
Documentation=https://docs.docker.com
After=network-online.target docker.socket firewalld.service
Wants=network-online.target
Requires=docker.socket
[Service]
Type=notify
# the default is not to use systemd for cgroups because the delegate issues still
# exists and systemd currently does not support the cgroup feature set required
# for containers run by docker
ExecStart=/usr/bin/dockerd -H fd://
ExecReload=/bin/kill -s HUP $MAINPID
LimitNOFILE=1048576
# Having non-zero Limit*s causes performance problems due to accounting overhead
# in the kernel. We recommend using cgroups to do container-local accounting.
LimitNPROC=infinity
LimitCORE=infinity
# Uncomment TasksMax if your systemd version supports it.
# Only systemd 226 and above support this version.
TasksMax=infinity
TimeoutStartSec=0
# set delegate yes so that systemd does not reset the cgroups of docker containers
Delegate=yes
# kill only the docker process, not all processes in the cgroup
KillMode=process
# restart the docker process if it exits prematurely
Restart=on-failure
StartLimitBurst=3
StartLimitInterval=60s
[Install]
WantedBy=multi-user.targetUpdate docker; Verify that it's upgraded, and started; |
|
placed docker.socket in /lib/systemd/system, still same error now i saw that /var/run/docker.sock is a folder?! |
|
@reloxx13: Please run systemctl deamon-reload on a debian stretch system after you placed the docker.socket file int the folder /lib/systemd/system. After that you can run systemctl restart docker.service |
|
nope, wont start :/ after some googling (moby/moby#30348) i moved the folder away and now docker starts again 👍 My workaround for Rasbian; check if file exists, only if exists check the content, has to be this:
if docker started, remove |
does docker-compose change anything ? |
I suspect you've (at some point) had a container that bind-mounted the docker-socket. In case of a race condition (where the docker socket is not yet up, but the container is), that will result in docker creating a directory at that location if you're using the To prevent that from happening; use the So instead of Use this; |
no, it shouldn't. docker-compose is a whole separate package |
|
@thaJeztah yep, portainer is using docker.sock, but it shows it uses bind and not volume
think due the failed startups and missing unix:// it just failed at some point. never had that directory msg before. |
|
If you're specifying a host path, it's a bind-mount, not a volume. It's confusing, because at the time the |
|
But it also depends which part of the API is used for those settings (the "mount" or "volume" API) 😅 |
|
Downgrading as mentioned by @Val fixed the issue for me - thank you.
|
|
Hello, I'm using My systemd unit file is same as https://raw.githubusercontent.com/docker/docker-ce/v18.09.0/components/packaging/systemd/docker.service I can't start docker service (after upgrade from 18.06 version), but mine fails on Edit: And I've found same open issue #475 |
Hi Sebastiaan, when will be the upcoming patch release or package update for 18.09.0 CE? Is it a monthly cadence? |
I think the goal is to have a release early/mid december, but there are some issues still being looked into that are desirable for inclusion, so the "safe" answer is "when it's done". There's a workaround for this issue though (see #476 (comment)) |
correct. Ubuntu 16.04: Running //edit: Seems like using the override-mechanism using |
This is in fact another workaround, which will help to successfully upgrade to 18.09. systemctl status docker Dec 05 15:32:15 docker-single-host systemd[1]: Starting Docker Application Container Engine... |
|
18.09.1 was released yesterday, so it should be there |
|
Getting similar errors starting the service. That is not to say I am not being stupid though. |
Seems to work fine for me once I enable/start the socket |
|
Can anyone confirm that CentOS stable works now from performing a simple |
|
did not receive on rasbian till now :S
|
|
after 4 months, still not possible to install and run docker on a fresh rasbian without a workaround. if needed i can grant ssh access to the devs. docker-socket was still missing |
|
On Debian 9: |
|
Can you check the logs for docker.service and containerd.service? Unfortunately the output of systemd is usually not showing the actual error (other than: it failed to start) |
|
@thaJeztah I looked into Rebooting the machine solved the issue. Thanks for answering on Saturday!) |
|
I have same issue, here is my solution based on @audioscavenger comment: sudo systemctl edit docker.service add: then reload and restart docker: |
|
@wildanmuhlis2 what version of docker do you have installed? no more changes should be needed in current versions (unless you want to customise where the API is listening on) |
|
@thaJeztah Ubuntu Version: |
|
Hello Getting the below errors Dec 22 05:58:49 bond systemd[1]: docker.service: Service hold-off time over |
|
docker version: |
|
@xll1452 that error in itself doesn't tell much;
|
|
I Googled for a while, only to find StackOverflow questions and Docker Forums posts with no answers. It looks like installing Docker CE on a fresh Ubuntu Server install (20.04 LTS) isn't working properly due to the same issue as here. I've checked, and the installed version seemed to be the latest stable available. The error prevents the whole Edit: it seems similar to #989, but it never was resolved |
|
are you running the install manually, or also automated, similar to the linked issue? |
|
I was trying to install Docker during a Packer build, which executes scripts via SSH, so I'd say it was also automated. |
|
Hm, possibly yes. I haven't seen the problem occur in manual installs. I did find some discussion around a similar issue with CUPS, but at a glance, they "solved" it by not installing socket activation 🤔 https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=747073 |
|
same problem here on fresh install ubuntu 20.04, sudo apt install docker-ce I have solved in my machine : I need to turn off my vpn first https://forums.docker.com/t/subprocess-installed-post-installation-script-returned-error-exit-status-1/37444/4 |
Same problem on fresh Ubuntu Server install (20.04 LTS) @ptrcnull did you manage to solve issue? I can work around it but its really annoying that containers wont restart on host machine reboot :( ! @thaJeztah not sure if my 2cents helps, I used a simple install script to add docker and other dependencies on host machine. |
|
@cgd1 I ended up installing |
Worked for me with no error. |
A thousand thank you's !! 🥇 |
|
interesting... I actually had docker.io installed. still experiencing issue |
and others
Expected behavior
Upgrade to 18.09.0
3-0debian-stretch works painlessley.Actual behavior
Upgrade to 18.09.0
3-0debian-stretch removes the file /lib/systemd/system/docker.socket and does not restart after the upgrade.Steps to reproduce the behavior
Run apt-get dist-upgrade on a Debian stretch system with an older docker-ce package installed.
The text was updated successfully, but these errors were encountered: