Systemd starts but services don't
pablorq opened this issue · comments
After several tests, I can't find out how to do for systemd starts all the services.
If I start it without entering bash, I've got this:
>wsl -l -v
NAME STATE VERSION
* Ubuntu-20.04 Stopped 2
docker-desktop-data Running 2
docker-desktop Running 2
>wsl /bin/false
Sleeping for 1 second to let systemd settle
>wsl ps -ef
UID PID PPID C STIME TTY TIME CMD
root 1 0 2 14:36 ? 00:00:00 /lib/systemd/systemd --unit=basic.target
root 46 1 2 14:36 ? 00:00:00 /lib/systemd/systemd-journald
root 63 1 2 14:36 ? 00:00:00 /lib/systemd/systemd-udevd
systemd+ 67 1 1 14:36 ? 00:00:00 /lib/systemd/systemd-networkd
root 240 1 0 14:36 ? 00:00:00 snapfuse /var/lib/snapd/snaps/lxd_18150.snap /snap/lxd/18150 -o ro,n
message+ 246 1 0 14:36 ? 00:00:00 /usr/bin/dbus-daemon --system --address=systemd: --nofork --nopidfil
root 247 1 1 14:36 ? 00:00:00 /usr/sbin/NetworkManager --no-daemon
root 255 1 0 14:36 ? 00:00:00 /lib/systemd/systemd-hostnamed
systemd+ 266 1 1 14:36 ? 00:00:00 /lib/systemd/systemd-resolved
root 316 63 2 14:36 ? 00:00:00 /lib/systemd/systemd-udevd
root 317 63 2 14:36 ? 00:00:00 /lib/systemd/systemd-udevd
root 318 63 0 14:36 ? 00:00:00 /lib/systemd/systemd-udevd
root 319 0 0 14:36 pts/0 00:00:00 /usr/bin/sudo -H -u [USER] /bin/bash -c set -a; [ -f "$HOME/.system
[USER] 320 319 0 14:36 pts/0 00:00:00 ps -ef
The /bin/false command is for trigger your script to enter the systemd namespace, without doing any else.
In case I enter the shell, I've got this:
>wsl -l -v
NAME STATE VERSION
* Ubuntu-20.04 Stopped 2
docker-desktop-data Running 2
docker-desktop Running 2
>wsl
Sleeping for 1 second to let systemd settle
Welcome to Ubuntu 20.04.1 LTS (GNU/Linux 4.19.128-microsoft-standard x86_64)
* Documentation: https://help.ubuntu.com
* Management: https://landscape.canonical.com
* Support: https://ubuntu.com/advantage
System information as of vie 27 nov 2020 14:37:24 CET
System load: 0.6 Processes: 39
Usage of /: 2.6% of 250.98GB Users logged in: 0
Memory usage: 6% IPv4 address for eth0: 192.168.26.186
Swap usage: 0%
* Introducing self-healing high availability clustering for MicroK8s!
Super simple, hardened and opinionated Kubernetes for production.
https://microk8s.io/high-availability
0 actualizaciones se pueden instalar inmediatamente.
0 de estas actualizaciones son una actualización de seguridad.
Last login: Fri Nov 27 14:37:06 CET 2020 on pts/0
$ ps -ef
UID PID PPID C STIME TTY TIME CMD
root 1 0 1 14:37 ? 00:00:00 /lib/systemd/systemd --unit=basic.target
root 46 1 0 14:37 ? 00:00:00 /lib/systemd/systemd-journald
root 63 1 0 14:37 ? 00:00:00 /lib/systemd/systemd-udevd
systemd+ 70 1 0 14:37 ? 00:00:00 /lib/systemd/systemd-networkd
root 180 0 0 14:37 pts/0 00:00:00 /bin/login -p -f 'HOSTTYPE=x86_64' 'PWD=/mnt/d/apps/System'
root 256 1 0 14:37 ? 00:00:00 snapfuse /var/lib/snapd/snaps/lxd_18150.snap /snap/lxd/18150 -o ro,n
message+ 262 1 0 14:37 ? 00:00:00 /usr/bin/dbus-daemon --system --address=systemd: --nofork --nopidfil
root 263 1 0 14:37 ? 00:00:00 /usr/sbin/NetworkManager --no-daemon
systemd+ 283 1 0 14:37 ? 00:00:00 /lib/systemd/systemd-resolved
root 366 1 0 14:37 ? 00:00:00 /lib/systemd/systemd-logind
[USER] 368 1 0 14:37 ? 00:00:00 /lib/systemd/systemd --user
[USER] 369 368 0 14:37 ? 00:00:00 (sd-pam)
[USER] 374 368 0 14:37 ? 00:00:00 /usr/bin/pulseaudio --daemonize=no --log-target=journal
[USER] 376 368 0 14:37 ? 00:00:00 /usr/libexec/tracker-miner-fs
[USER] 377 180 0 14:37 pts/0 00:00:00 -bash
[USER] 385 368 0 14:37 ? 00:00:00 /usr/bin/dbus-daemon --session --address=systemd: --nofork --nopidfi
rtkit 394 1 0 14:37 ? 00:00:00 /usr/libexec/rtkit-daemon
root 398 1 0 14:37 ? 00:00:00 /usr/lib/policykit-1/polkitd --no-debug
[USER] 406 368 0 14:37 ? 00:00:00 /usr/libexec/gvfsd
[USER] 421 368 0 14:37 ? 00:00:00 /usr/libexec/gvfsd-fuse /run/user/1000/gvfs -f -o big_writes
[USER] 427 368 0 14:37 ? 00:00:00 /usr/libexec/gvfs-udisks2-volume-monitor
root 431 1 0 14:37 ? 00:00:00 /usr/lib/udisks2/udisksd
[USER] 456 368 0 14:37 ? 00:00:00 /usr/libexec/gvfs-goa-volume-monitor
[USER] 460 368 0 14:37 ? 00:00:00 /usr/libexec/goa-daemon
[USER] 467 368 0 14:37 ? 00:00:00 /usr/libexec/goa-identity-service
[USER] 473 368 0 14:37 ? 00:00:00 /usr/libexec/gvfs-afc-volume-monitor
[USER] 478 368 0 14:37 ? 00:00:00 /usr/libexec/gvfs-mtp-volume-monitor
[USER] 482 368 0 14:37 ? 00:00:00 /usr/libexec/gvfs-gphoto2-volume-monitor
root 487 1 0 14:37 ? 00:00:00 /usr/lib/upower/upowerd
[USER] 513 377 0 14:37 pts/0 00:00:00 ps -ef
In any case I can start the services manually, but not all of them. For example:
>wsl -l -v
NAME STATE VERSION
* Ubuntu-20.04 Stopped 2
docker-desktop-data Running 2
docker-desktop Running 2
>wsl /bin/false
Sleeping for 1 second to let systemd settle
>wsl ps -ef
UID PID PPID C STIME TTY TIME CMD
root 1 0 1 14:41 ? 00:00:00 /lib/systemd/systemd --unit=basic.target
root 46 1 1 14:41 ? 00:00:00 /lib/systemd/systemd-journald
root 63 1 1 14:41 ? 00:00:00 /lib/systemd/systemd-udevd
systemd+ 70 1 0 14:41 ? 00:00:00 /lib/systemd/systemd-networkd
root 242 1 0 14:41 ? 00:00:00 snapfuse /var/lib/snapd/snaps/lxd_18150.snap /snap/lxd/18150 -o ro,n
message+ 248 1 0 14:41 ? 00:00:00 /usr/bin/dbus-daemon --system --address=systemd: --nofork --nopidfil
root 249 1 0 14:41 ? 00:00:00 /usr/sbin/NetworkManager --no-daemon
root 254 1 0 14:41 ? 00:00:00 /lib/systemd/systemd-hostnamed
root 258 1 0 14:41 ? 00:00:00 /usr/lib/NetworkManager/nm-dispatcher
systemd+ 269 1 0 14:41 ? 00:00:00 /lib/systemd/systemd-resolved
root 322 63 0 14:41 ? 00:00:00 /lib/systemd/systemd-udevd
root 323 63 0 14:41 ? 00:00:00 /lib/systemd/systemd-udevd
root 324 63 0 14:41 ? 00:00:00 /lib/systemd/systemd-udevd
root 325 63 0 14:41 ? 00:00:00 /lib/systemd/systemd-udevd
root 327 63 1 14:41 ? 00:00:00 /lib/systemd/systemd-udevd
root 330 0 0 14:41 pts/0 00:00:00 /usr/bin/sudo -H -u [USER] /bin/bash -c set -a; [ -f "$HOME/.system
[USER] 331 330 0 14:41 pts/0 00:00:00 ps -ef
>wsl sudo systemctl start cron.service
[sudo] password for [USER]:
>wsl ps -ef
UID PID PPID C STIME TTY TIME CMD
root 1 0 0 14:41 ? 00:00:00 /lib/systemd/systemd --unit=basic.target
root 46 1 0 14:41 ? 00:00:00 /lib/systemd/systemd-journald
root 63 1 0 14:41 ? 00:00:00 /lib/systemd/systemd-udevd
systemd+ 70 1 0 14:41 ? 00:00:00 /lib/systemd/systemd-networkd
root 242 1 0 14:41 ? 00:00:00 snapfuse /var/lib/snapd/snaps/lxd_18150.snap /snap/lxd/18150 -o ro,n
message+ 248 1 0 14:41 ? 00:00:00 /usr/bin/dbus-daemon --system --address=systemd: --nofork --nopidfil
root 249 1 0 14:41 ? 00:00:00 /usr/sbin/NetworkManager --no-daemon
systemd+ 269 1 0 14:41 ? 00:00:00 /lib/systemd/systemd-resolved
root 339 63 0 14:41 ? 00:00:00 /lib/systemd/systemd-udevd
root 340 63 0 14:41 ? 00:00:00 /lib/systemd/systemd-udevd
root 341 63 0 14:41 ? 00:00:00 /lib/systemd/systemd-udevd
root 342 63 0 14:41 ? 00:00:00 /lib/systemd/systemd-udevd
root 347 1 0 14:41 ? 00:00:00 /usr/sbin/cron -f
root 348 0 0 14:42 pts/0 00:00:00 /usr/bin/sudo -H -u [USER] /bin/bash -c set -a; [ -f "$HOME/.system
[USER] 349 348 0 14:42 pts/0 00:00:00 ps -ef
Any idea how to do for starting all the missing services?
[SOLVED] All enabled systemd services are run on boot
Doing a lot of research I was thinking about to read all enabled services and start all of them with an script, after systemd was running as PID 1.
But this is the work that systemd has to do... when going to multi-user.target.
So that's all. Change from --unit=basic.target to --unit=multi-user.target in both scripts and reboot wsl (no the windows machine). After that you will have all services running, like cron or snap. In fact, this fix the "all snap packages broken after boot" issue.
Summary
Changes to do:
File start-systemd-namespace
#SYSTEMD_EXE="/lib/systemd/systemd --unit=basic.target"
SYSTEMD_EXE="/lib/systemd/systemd --unit=multi-user.target"
File enter-systemd-namespace
#SYSTEMD_EXE="/lib/systemd/systemd --unit=basic.target"
SYSTEMD_EXE="/lib/systemd/systemd --unit=multi-user.target"
WARINING
A collateral effect is related to DNS resolution, and the WSL original name resolution file is overwritten by a new one with wrong nameservers.
Easy fix: disable systemd name resolution:
$ sudo systemctl disable systemd-resolved.service