systemd pauses for unspecified reason

[ralphb]

I recently installed Kubuntu 20.04 on a new Ryzen 3900X system. I noticed that the boot time (time between entering LUKS credentials and appearance of login screen) took more than 1 minute.

Using `systemd-analyze`, I found the following:

```
cassiopeia:~# systemd-analyze
Startup finished in 24.819s (kernel) + 1min 32.150s (userspace) = 1min 56.969s
graphical.target reached after 1min 32.144s in userspace
cassiopeia:~# systemd-analyze blame
4.721s fstrim.service
2.949s apt-daily-upgrade.service
2.380s <email address hidden>
1.739s NetworkManager-wait-online.service
 439ms apt-daily.service
 421ms man-db.service
 398ms systemd-logind.service
 323ms dev-mapper-root.device
 294ms <email address hidden>
 204ms logrotate.service
 153ms systemd-journald.service
...
cassiopeia:~# systemd-analyze critical-chain
The time when unit became active or started is printed after the "@" character.
The time the unit took to start is printed after the "+" character.

graphical.target @1min 32.144s
└─multi-user.target @1min 32.144s
  └─fetchmail.service @1min 32.128s +16ms
    └─network-online.target @1min 32.127s
      └─NetworkManager-wait-online.service @1min 30.388s +1.739s
        └─NetworkManager.service @1min 30.282s +105ms
          └─dbus.service @1min 30.279s
            └─basic.target @1min 30.265s
              └─sockets.target @1min 30.265s
                └─snapd.socket @1min 30.264s +370us
                  └─sysinit.target @1min 30.262s
                    └─systemd-timesyncd.service @1min 30.219s +42ms
                      └─systemd-tmpfiles-setup.service @1min 30.202s +15ms
                        └─systemd-journal-flush.service @288ms +49ms
                          └─systemd-journald.service @131ms +153ms
                            └─systemd-journald.socket @129ms
                              └─-.mount @127ms
                                └─<email address hidden> @410ms
                                  └─<email address hidden> @401ms +8ms
                                    └─dev-nvme0n1p2.device @398ms
```

This didn't really help, but `systemd-analyze plot > boot.svg` produced an SVG file (see attachment) that shows a gap between store.mount (completes within 38 ms) and apparmor.service of almost 1:30 minutes. What did systemd do during this gap?

I think it is either a bug in systemd that idles, or in systemd-analyze that doesn't disclose the actions in that gap.

(Note: I have posted this question in two well-known Ubuntu forums, but I didn't receive any answers.)

[Dan Streetman (ddstreet)]

you should check your journal to see what service timed out, as that is most likely what caused the delay.

[ralphb (dev-endlos)]

The issue was indeed /etc/fstab, which contained a UUID inserted by the installer. Removing the UUID by the actual device name removed the 1:30 minutes delay.

You might argue that this is a bug by itself, but it's good enough for me.