Bugzilla – Bug 635851
ntp, dovecot/kdm: programs exit prematurely (or don't start at all) + X server timeout
Last modified: 2015-11-06 09:02:00 UTC
Hi, I have here a diskless system where kdm runs into believing that the X server ran into a startup timeout. When kdm starts the X server, the screen resolution does change, I see the KDM busy mouse cursor for one or two seconds, after which the X server gets terminated again. Please advise.
Created attachment 386499 [details] /var/log/messages PID TTY STAT TIME COMMAND 2268 ? Ss 0:00 \_ sshd: root@pts/0 2272 pts/0 Ss 0:00 \_ -bash 2340 pts/0 S+ 0:00 \_ w3m bugzilla.novell.com 2354 pts/0 S+ 0:00 \_ /usr/bin/joe /root/.w3m/w3mtmp2340-3 2355 pts/0 S+ 0:00 \_ /usr/bin/joe /root/.w3m/w3mtmp234 2356 pts/0 S+ 0:00 \_ /usr/bin/joe /root/.w3m/w3mtmp234 2357 pts/1 Rs+ 0:00 \_ ps fx 2000 ? S 0:00 /usr/bin/kdm 2179 ? Ssl 0:00 /usr/sbin/automount -p /var/run/automount.pid 2249 tty1 Ss+ 0:00 /sbin/mingetty --noclear tty1 2250 tty2 Ss+ 0:00 /sbin/mingetty tty2 2251 tty3 Ss+ 0:00 /sbin/mingetty tty3 2252 tty4 Ss+ 0:00 /sbin/mingetty tty4 2253 tty5 Ss+ 0:00 /sbin/mingetty tty5 2254 tty6 Ss+ 0:00 /sbin/mingetty tty6 2260 ? Ss 0:00 /usr/sbin/sssd -f -D 2262 ? S 0:00 \_ /usr/lib/sssd/sssd_be -d 0 --debug-to-files --dom 2263 ? S 0:00 \_ /usr/lib/sssd/sssd_nss -d 0 --debug-to-files 2264 ? S 0:00 \_ /usr/lib/sssd/sssd_pam -d 0 --debug-to-files That is really all what "ps fx" has to say about kdm. Restarting kdm by means of loggin in and `rcxdm restart` does work, that is, kdm will show me the login window and not die.
/var/log/kdm.log /var/log/X*
Created attachment 386867 [details] kdm.log
Created attachment 386868 [details] Xorg.0.log
I just had this occur with a workstation and discovered the issue is as follows: System BIOS is set to local time System boots and expects system clock is set to UTC kdm starts a timeout (120sec?) on X NTP gets a time update from the server Time now jumps by 5 or 6 hours, hence KDM declares that X has timed out My solution was to uncheck the "Hardware clock set to UTC" in Yast2 "Date and Time" and ensure that the BIOS clock was reasonable. You might want to check if this is also your issue. It would be nice to have a solution from the openSUSE team, perhaps allowing a delay on the startup of the ntp client, have kdm wait for ntp to get a refresh, or even have kdm try the timeout twice.
Hi there, I just experienced this BUG today on about 20 OpenSuse 11.4 machines: Last weekend, we had the change from winter to summer time in Germany, i.e. all BIOS clocks show 08:00 AM when our NTP server says its 09:00 AM. Guess what: All 20 machines fail to start the X server with: --------- kdm: X server startup timeout, terminating kdm: X server for display :0 cannot be started, session disabled --------- My current is setting in Yast is that "Hardware clock set to UTC" is DISABLED. If I enable the checkbox, the X server starts just fine. Still, this should not happen like this... I changed the product to 11.4!
Well, that's a feature in the display managers to no longer try to start X again and again if it fails 3 times in a row in a short time period. AFAIK this is implemented via comparing timestamps when X startup begins and when it's available for X clients. Apparently the timestamp of the latter is measured as before the one for the former. Is this also reproducable with xdm? Then the X guys can have a lookt at this.
I just tested a couple of times: With - KDM the X startup fails, - XDM it starts fine. Is there any logfile I could provide to help debugging?
Not sure, which logfile would be useful here. Better let's reassign it to KDE component.
Hi Stefan, thanks for your hint about timestamps. I had the same problem as Joschi with some openSUSE 11.3 machines since March 28, which was the first Monday after changing to European summer time. kdm and ntpd conflicted, so the x-server thought it was working already for one hour. I deleted "vga=3..." in "/boot/grub/menu.lst" and booting to runlevel 5 works fine now (but without splash screen). Changing the order of init.d-jobs is a bit more effort, if you love to have the splash screen, but works, too.
Just a quick question: did removing the vga option help? or did you rearrange to order of init.d jobs? which one solved the problem for you? Thanks, Joschi
Sorry, Joschi, I was to quick responding. It worked on my machine, but not on the other machines. Finally I included "hwclock --systohc" to /etc/init.d/ntp to set the hardware clock to MESZ/CEST after NTP-synchronization. It's a workaround, but it helps. You have to restart ntp daemon or boot the machine twice.
Setting the clock is the problem. It does not just affect kdm, it also affects dovecot as I have recently noticed. Dovecot especially does not like when the time is stepped backwards, causing dovecot to exit.
The clocks changed in my time zone on Sunday, and since then I'm experiencing the same problem: the UI fails to start on bootup, dropping me into the text console. If I log in as root and issue rcxdm restart, then KDM starts without any error. This bug makes it impossible for anyone without root access to start the machine. Notker, where in /etc/init.d/ntp did you add the "hwclock --systohc" line? I'd like to try this workaround but it's not obvious to me where to apply it. I'm available to help troubleshooting or testing fixes.
We now use sddm, kdm is not actively developed. I will close. If this bug still applies for you please reopen.