Installation not working. Update to #55

[ofirule]

Installation process I went through on ubuntu 16.04 arch: arm64

Based on: https://github.com/balena-os/balena-engine/issues/55#issuecomment-423890806

1 run install script curl -sfL https://balena.io/engine/install.sh | sh

2 copy balena.service balena.socket from : https://github.com/balena-os/balena-engine/tree/17.06-resin/contrib/init/systemd to directory: /etc/systemd/system

3 create group balena-engine sudo groupadd balena-engine

4 update /etc/systemd/system/balena.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=balena-engine.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/local/bin/balena-engine-daemon
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.target

changes: ExecStart - 1) change executable path since install.sh puts it in different path 2) remove -H flag since it causes an error TasksMax - Uncommented TimeoutStartSec - Commented, it causes an error

Please update install.sh and or service configuration

[robertgzr]

Thanks for this detailed report.

I have started working on a PR that fixes this and clears up some other things here #129

[robertgzr]

Actually @ofirule can you send a PR with the service tweaks you did? I would like to restrict my PR to only changes to the install script itself...

[ofirule]

Creating PR :+1:

[rgaudin]

FYI this is still not up to date with what's provided in install.sh.

Here's a configuration that works now, on raspbian 2021-05-07:

curl -sfL https://balena.io/engine/install.sh | sh
sudo groupadd balena-engine
# Add files listed bellow
sudo systemctl daemon-reload
sudo systemctl start balena
sudo systemctl enable balena  # start on boot
sudo usermod -aG balena-engine pi  # (or whatever user) to enable non-root balena mgmt

/etc/systemd/system/balena.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=balena.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/local/bin/balena-engine-daemon
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.target

/etc/systemd/system/balena.socket

[Unit]
Description=Docker Socket for the API
PartOf=balena.service

[Socket]
ListenStream=/var/run/balena-engine.sock
SocketMode=0660
SocketUser=root
SocketGroup=balena-engine

[Install]
WantedBy=sockets.target

[bergpb]

Hello @rgaudin, I'm trying to install the balena-engine in a Raspberry Pi Zero with Raspbian and for this, I've used the steps described above by you, but it does not work. Checking logs wth journalctl -u balena -f if was found this information:

May 11 10:04:11 anna systemd[1]: balena.service: Scheduled restart job, restart counter is at 1.
May 11 10:04:11 anna systemd[1]: Stopped Docker Application Container Engine.
May 11 10:04:11 anna systemd[1]: Starting Docker Application Container Engine...
May 11 10:04:11 anna systemd[1]: balena.service: Main process exited, code=dumped, status=4/ILL
May 11 10:04:11 anna systemd[1]: balena.service: Failed with result 'core-dump'.
May 11 10:04:11 anna systemd[1]: Failed to start Docker Application Container Engine.
May 11 10:04:12 anna systemd[1]: balena.service: Scheduled restart job, restart counter is at 2.
May 11 10:04:12 anna systemd[1]: Stopped Docker Application Container Engine.
May 11 10:04:12 anna systemd[1]: Starting Docker Application Container Engine...
May 11 10:04:12 anna systemd[1]: balena.service: Main process exited, code=dumped, status=4/ILL
May 11 10:04:12 anna systemd[1]: balena.service: Failed with result 'core-dump'.
May 11 10:04:12 anna systemd[1]: Failed to start Docker Application Container Engine.
May 11 10:04:12 anna systemd[1]: balena.service: Scheduled restart job, restart counter is at 3.

[rgaudin]

@bergpb my working code is at https://github.com/rgaudin/bard-image/blob/main/prepare-system.sh#L53 There are a few things (that I forgot) to care about. Like the docker path symlink (I think it was for compose) or the binary name (what the binary does is switched based on called name).

[bergpb]

Nice thanks @rgaudin, after following your installation steps I was found another error on the start balena service:

May 11 11:02:07 anna balena-engine-daemon[6972]: Error starting daemon: Devices cgroup isn't mounted

Do I need to add anything in boot.txt to make it works?

[rgaudin]

There's been a change in cgroups layout in debian buster and balena hasn't been updated to support it so you'd need to stick to 2021-12-02-raspios-buster-armhf-lite until this is handled. I suppose raspian docker is up to date but haven't used it

[bergpb]

Checking I already have the cgroup mounted as expected:

pi@anna:~ $ mount | grep cgroup
cgroup2 on /sys/fs/cgroup type cgroup2 (rw,nosuid,nodev,noexec,relatime,nsdelegate,memory_recursiveprot)

But looks like the installation only supports the cgroups v1: https://github.com/balena-os/balena-engine/issues/284

Very appreciate your help @rgaudin I will try to downgrade to the version 2021-12-02-raspios-buster-armhf-lite and use it, until wait for a new release for balena-engine with support to cgroups2

[bergpb]

Thanks @rgaudin, after install raspbian version 2021-12-02-raspios-buster-armhf-lite and made some adaptations with your install script works fine. Also posting gist here, can help another person in the future: https://gist.github.com/bergpb/50340bcb1599079a23d17fd9d7237e9d