pdist: Add pdistd service for distmaster(s) #326
Conversation
|
s/out-update/auto-update/ |
There was a problem hiding this comment.
telnet deinemuddah 237 von maleficent führt zu den folgenden Zeilen in journalctl auf deinemuddah:
Jun 13 15:56:43 deinemuddah.molgen.mpg.de systemd[1]: Started 141.14.18.119:55986.
Jun 13 15:57:30 deinemuddah.molgen.mpg.de systemd[1]: pdistd@7-141.14.30.156:237-141.14.18.119:55986.service: Succeeded.
Jun 13 15:57:30 deinemuddah.molgen.mpg.de systemd[1]: pdistd@7-141.14.30.156:237-141.14.18.119:55986.service: Consumed 14.824s CPU time.
|
Fixed Problem with pdist-bootcheck: The pcpool systems stop console output after i915 drm initialization. But if we do syncrchonous pdist update and reboot without any indication to the user what is happening, the user will assume the system to be frozen. I guess that is the same problem, why the virtual consoles are not working there. |
|
For some reason blacklist i915 in modprobe.d doesn't work. |
|
Close was accidentally. |
We have some smaller systems with integrated Intel graphics, for example amaru and machdoch. |
|
Maybe it’s the following issues:
A fix could be commit https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=5ae3716cfdcd286268133867f67d0803847acefc: |
|
Jun 16 12:34:18 icebear.molgen.mpg.de systemd[1]: named.service: Found ordering cycle on automount.service/start |
|
In pdist> Add pdist-bootcheck do you mean a colon :? |
Yeah, but the rest of the commit message is also not final :-) |
With this patch or with a 6.3 kernel, virtual consoles work again. However,the console output is still lost during boot when the i915 driver initialized, presumably because the output is then going to the video output of the intel card. This can not be confirmed, because when a monitor is connected to the intel hdmi output, it is used right away. |
|
Maybe it not because of the 6.3 kernel that the console comes back after the time, but because of the fact we didn't have a Intel driver for the test kernel and so the nouveau was loaded instead. The nouveau driver seems to have the functionality needed for the virtual consoles. But still the display is gone when i915 starts and comes back seconds later with a cleared screen when noveau starts. Usually you wouldn't notice, because at the same time the display manage starts and clears the screen again. Idea: Avoid autoloading i915 and continue with the initial console (configure by bios, I guess) for as long as possible. This way, we can see the boot messages and errors and are able to notify the user if a update is required and causes longer delays. When that is all done and we are ready to switch to graphic mode, manually load i915 manually just as we load nouveau or Nvidia. |
donald
force-pushed
the
add-pdistd
branch
3 times, most recently
from
88e3ef7
to
d8508af
Compare
|
This in on top of #328 |
To implement auto-update on boot, we need a way for a client to trigger a pdist-update for itself on its distmaster. Add a service for the distmaster(s), which allows a client to do that by connecting to a fixed port on the distmaster (`telnet deinemuddah 237`).
Add service to check pdist time for the booting system. If older than 7 days, request a synchonous pdist from the distmater and reboot. If only older than 1 day, require an asynchronous pdist while the boot can continue.
Add a network service on the distmaster for clients to use to request a pdist
Add a startup service for desktops to check their pdist status and request either an asynchronous pdist update or an synchronous pdist update followed by a reboot if needed.