I'm encountering this shutdown delay too frequently. What causes it? Current instance is on Buster, host fi965.
On Wed July 24 2019 09:41:18 Felix Miata wrote:
I'm encountering this shutdown delay too frequently. What causes it? Current instance is on Buster, host fi965.
FWIW I'm not seeing that with several sysvinit Busters - mostly amd64 or multiarch but also at least one very very slow 192MB i386 test box.
--Mike
On Wed, Jul 24, 2019 at 12:57 PM Mike Bird mgb-trinity@yosemite.net wrote:
On Wed July 24 2019 09:41:18 Felix Miata wrote:
I'm encountering this shutdown delay too frequently. What causes it? Current instance is on Buster, host fi965.
FWIW I'm not seeing that with several sysvinit Busters - mostly amd64 or multiarch but also at least one very very slow 192MB i386 test box.
--Mike
Hi Mike the stop jobs are almost always caused by systemd services so you should not get them on systems using sysvinit. Tracking them down is sometimes a real pain the other option is to change the time out on systemd to force it to terminate sonner than 90 seconds.
Anno domini 2019 Wed, 24 Jul 09:57:40 -0700 Mike Bird scripsit:
On Wed July 24 2019 09:41:18 Felix Miata wrote:
I'm encountering this shutdown delay too frequently. What causes it? Current instance is on Buster, host fi965.
FWIW I'm not seeing that with several sysvinit Busters - mostly amd64 or multiarch but also at least one very very slow 192MB i386 test box.
--Mike
To unsubscribe, e-mail: trinity-users-unsubscribe@lists.pearsoncomputing.net For additional commands, e-mail: trinity-users-help@lists.pearsoncomputing.net Read list messages on the web archive: http://trinity-users.pearsoncomputing.net/ Please remember not to top-post: http://trinity.pearsoncomputing.net/mailing_lists/#top-posting
I have seen it, too, on beowulf and ascii. My workaround was just to kill tde the hard way when I shut down my computers (over acpi hooks)
Nik
Mike Bird wrote:
FWIW I'm not seeing that with several sysvinit Busters - mostly amd64 or multiarch but also at least one very very slow 192MB i386 test box.
It is a known bug - I do not recall which one, but the essence is that in
/lib/systemd/system/tdm.service
you should remove
Conflicts=getty@tty7.service plymouth-quit.service
or just comment it out. Then systemd will shutdown tdm properly
regards
deloptes composed on 2019-07-24 20:00 (UTC+0200):
Mike Bird wrote:
FWIW I'm not seeing that with several sysvinit Busters - mostly amd64 or multiarch but also at least one very very slow 192MB i386 test box.
It is a known bug -
Trinity bug? Systemd? Debian?
# grep RETT /etc/os-release PRETTY_NAME="Debian GNU/Linux bullseye/sid" # dpkg -l | grep outh # systemctl list-unit-files | egrep 'outh|tty7' # systemctl list-units | egrep 'outh|tty7' # grep outh /lib/systemd/system/tdm.service Conflicts=getty@tty7.service plymouth-quit.service After=systemd-user-sessions.service getty@tty7.service plymouth-quit.service #
Nothing should ever wait on something that does not exist.
I do not recall which one, but the essence is that in
/lib/systemd/system/tdm.service
you should remove
Conflicts=getty@tty7.service plymouth-quit.service
or just comment it out. Then systemd will shutdown tdm properly
Instead, I tried two versions of /etc/systemd/system/tdm.service, both of which worked:
First try: ********** [Unit] Description=Trinity Display Manager Documentation=man:tdm-trinity(1) Conflicts= After=
[Service] # temporary safety check until all DMs are converted to correct # display-manager.service symlink handling ExecStartPre=/bin/sh -c '[ "$(basename $(cat /etc/X11/default-display-manager 2>/dev/null))" = "tdm" ]' ExecStart=/opt/trinity/bin/tdm Restart=always
Second try: *********** [Unit] Conflicts= After=
[Service]
On Wednesday 24 July 2019 20:53:57 Felix Miata wrote:
/etc/systemd/system/tdm.service
If I change to that :
Conflicts= After=
The X server doesn't start.
If I remove this line : "Conflicts=getty@tty7.service plymouth-quit.service" It seems better, no wait when I stop the computer.
Have a good day,
André
andre_debian@numericable.fr composed on 2019-07-25 10:43 (UTC+0200):
Felix Miata wrote:
/etc/systemd/system/tdm.service
If I change to that :
Conflicts= After=
The X server doesn't start.
If I remove this line : "Conflicts=getty@tty7.service plymouth-quit.service" It seems better, no wait when I stop the computer.
Whether what I did might work for others may depend on the distro used and/or whether or how plymouth is configured and/or installed. None of my installations have plymouth installed except for Mageia, which is unusable if plymouth installation is disallowed. In Mageia I include plymouth.enable=0 or noplymouth on Grub's kernel cmdline and/or disable plymouth with systemctl.
For purposes of this thread, /etc/systemd/system/tdm.service did not exist. I copied it from /lib/systemd/system/ and then modified it. IIRC, I was using Bullseye only, not Buster as I wrote in OP.
Felix Miata wrote:
Whether what I did might work for others may depend on the distro used and/or whether or how plymouth is configured and/or installed. None of my installations have plymouth installed except for Mageia, which is unusable if plymouth installation is disallowed. In Mageia I include plymouth.enable=0 or noplymouth on Grub's kernel cmdline and/or disable plymouth with systemctl.
For purposes of this thread, /etc/systemd/system/tdm.service did not exist. I copied it from /lib/systemd/system/ and then modified it. IIRC, I was using Bullseye only, not Buster as I wrote in OP.
Always better to test on default setup - I am not sure if plymouth is installed by default, but I tested on a fresh installed Debian and this exactly solved the problem. It was also confirmed in the original bug report. No need to further muddy the waters.
What I understand is perhaps it is a plymouth related. If I have the time I would do some testing, find the bug and update there.
regards
Felix Miata wrote:
Instead, I tried two versions of /etc/systemd/system/tdm.service, both of which worked:
It is enough to remove the conflicts for now - there will be a patch at some point of time in the future. The After part is OK and IMO it is important as it says when this is to be executed - obviously it contradicts to Conflicts and systemd stays waiting for something that never happens. But thank you for your testing - if you find the bug put it there.
Yes, I am on debian.
regards