[CentOS-devel] Problem with initial-setup-text systemd job on headless server
bertnz at gmail.com
Thu Oct 9 01:37:51 UTC 2014
If my email is TL;DR then the jist of my question/issue is:
On a headless machine, you are never presented with the prompt to set up a
user or accept the license agreement. (initial-setup job/package).
I do run VNC server, and had gone through the GNOME welcome wizard, as a
standard user and as root.
A systemd job keeps running on boot that's presumably outputting the
initial-setup-text program to [somewhere] but there's no way (that I know
of) to give it input to tell it to accept the agreement and continue.
The initial-setup-text systemd job means systemd-analyze outputs "Bootup is
not yet finished. Please try again later."
Manually running initial-setup-text and accepting the license agreement
didn't seem to affect any change. Running the program again just said the
license agreement hadn't been accepted.
I just removed the initial-setup package via yum.
The systemd job went away and systemd-analyze now reports bootup in x
Are there any unforeseeable ramifications of doing this? At least my
machine isn't spontaneously rebooting...
Does the initial-setup package change/mark something somewhere that should
perhaps be done manually?
More detailed version:
Following a test upgrade from 6.5 to 7 on a headless server (worked fine
btw) I ran into a bit of a SNAFU with systemd and services. I was trying to
troubleshoot a couple of legacy init.d scripts for some programs that just
wouldn't seem to start correctly via systemd using the chkconfig legacy
The systemd-analyze tool kept saying that my machine hadn't finished
I had performed the upgrade mid-September.
The server has been rebooted a few times, but even after 3 days, I still
got that message.
BS I said.
I've read the docs thoroughly on how to manage services, and, well, no
services were really causing any problems (a couple were failing on start
up, resolved most of those, still stuck on one but I could disable that) so
I couldn't understand why systemd-analyze would keep reporting that my
system hadn't "finished booting".
I then stumbled upon the "systemctl list-jobs".
This listed - among a couple of other jobs - an initial-setup-text program
running as a job that was "waiting" (from memory). Possibly waiting on
input. I can't remember.
Realizing what this was, I tried to run it manually from my ssh console.
Press 2, accepting the license agreement, c, c ... the job wouldn't "go
away". The license also was never accepted. I ran it again, accepted the
agreement, c, went back again. Still no change.
So I killed the job...
[root at trixie bin]# systemctl status initial-setup-text
initial-setup-text.service - Initial Setup configuration program (text mode)
Loaded: loaded (/usr/lib/systemd/system/initial-setup-text.service;
Active: active (exited) since Thu 2014-10-09 13:35:50 NZDT; 7min ago
Main PID: 993 (code=killed, signal=TERM)
..... and re-booted.
It came back (the job that is). FFS. Still couldn't really use
systemd-analyze to try and find out what's going on with my "legacy
services" that aren't playing nice.
yum remove initial-setup
... and now the job has gone. YAY. Maybe I should have done something
differently, Oh well.
My systemd-analyze now reports my system boots OK, and systemd list-jobs
reports no jobs running.
So I guess my bug report would go something like....
The initial-setup-text systemd job keeps coming back after reboots.
There's no way (easy way that I could find) to provide input to the
initial-setup-text systemd job on a headless system. Short of redirecting
output to a com port and ... yeah... too hard.
systemd never reports a successful boot if you have a job that never
completes (and that keeps coming back).
Removing the initial-setup-text package was the only way I could think of
to get rid of it.
My system doesn't behave erratically after doing this.
I would suggest (if possible) that the initial-setup-text gets run when you
SSH into your machine for the first time for you to accept the license
Maybe it's not even required. I did see a couple of bugzilla reports (
http://bugs.centos.org/view.php?id=7173) that maybe suggested that.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the CentOS-devel