Description of problem: runlevel unknown, no network after first "urpmi --auto-update" Version-Release number of selected component (if applicable): net-tools-2.0-0.20140707git.4.mga5 (I think it's the init scripts) How reproducible: I've installed Cauldron R2 as server (package-choice: minimal (with grumpi (2nd menupoint in installation)) without dokumentation (about 700MB)). - installed sshd, shorewall - reboot => runlevel 3, network-up, access per ssh perfect Then I did a "urpmi --auto-update" directly after installation. - reboot => runlevel unknown, no network, no access per ssh Logged-in as root - type "init 3" => runlevel 3, network-up, access per ssh perfect Steps to Reproduce: 1. repeat "How reproducible:" Reproducible: Steps to Reproduce:
I'm not sure I understand what you mean by "runlevel unknown", do you mean it ended up in rescue mode? Can you please, as root, run: journalctl -b > journalctl-b.txt and attach journalctl-b.txt to this bug report?
CC: (none) => marja11Component: Backports => RPM Packages
Whiteboard: (none) => NEEDINFO
Created attachment 5419 [details] urpmi --auto-update (console output)
Created attachment 5420 [details] jounaltctl -b directly after reboot
Created attachment 5421 [details] jounalctl -b after run: init 3
WOW!! fast reaction! by "runlevel unknown" I mean: - restart server - login as "root" - run: "runlevel" - runlevel throws "unknown" to console (network unreachable) - run "init 3" - run: "runlevel" - runlevel throws "N 3" to console (network is up) so it's not possible to update and reboot the server remote, because after reboot, I have to type "init 3" manually. After "init 3" is done, everythings running fine
p.s. in "jounaltctl-b_runlevel_unknown.txt" is "network-up" missing.
Thx for all the information. CC'ing Colin (who knows all about initscripts and systemd) and Thomas (who was the last to push net-tools)
CC: (none) => mageia, thomasWhiteboard: NEEDINFO => (none)
What does net-tools have to do with any of this? Also, what does this show? ls -l /etc/systemd/system/default.target
Hi David sorry about my mentioned "net-tools", but to create a bug, you have to choose a package. For I don't know where the bug is, I chose this one (seems plausible to me, that missing network have sth. to do with a package with network-stuff). But as I wrote in my first post "(I think it's the init scripts)", because network runs, if running "init 3" manually. ls -l /etc/systemd/system/default.target points to /usr/lib/system.d/system/multi-user.target I forgot, sth. to say: As I setup the server, the was a "small" update first (5/6 packages including clib). After this, the system is telling "please restart couse of new clib". The the second "big" update (see attachment 5419 [details]) was done behind. If you want to, I can setup the server again and do a - journalctl -b > journalctl-b.txt between "small" and "big" update. Or is the infos' I sent you enough to reproduce the bug?
Hi Marja, it seems to be a shorewall problem. I reinstalled the server: starts up and displays: > started IP-tables for firewall -> OK run "runlevel" => "N 3" reboot after the update: > xt_addrtype ipv6 does not support broadcast matching run "runlevel" => unknown run "init 3" run "runlevel" => N 3 after first installation I did a > urpmi.removemedia -a > urpmi.addmedia --distrib --mirror-list > urpmi.removemedia (all 32bit) > urpmi --auto-update - aria2 - glibc - lib64archive13 - lib64rpm3 - locales - locales-de - meta-task - perl-MDV-Distribconf - perl-URPM - rpm - urpmi reboot => till here everything is OK after 2nd installation (342 packages) the error occures.
Same here. Since some update via --auto-update the network not starting (correct) anymore after reboot. Same here, tell me that runlevel is "unknown".
CC: (none) => identity.mageia.org
Sounds to me like systemd isn't correctly setting the runlevel in utmp anymore.
Assignee: bugsquad => mageiaSource RPM: net-tools-2.0-0.20140707git.4.mga5 => systemd
Maybe (my knowledge isn't in this enough to discuss). I only know how to reproduce this but easily: - install minimum installation (ca: 750GB) - 1. update - 2. update - Error occures
One of my cauldron 64 bit systems, Nvidia, fail to boot with latest kernel 3.17.2-desktop-3.mga5 The boot messages on screen stops at " Starting Update UTMP about System Runlevel Changes" I also experience Bug 14457 - Missing hwdb.bin at start of, boot me at #5 I can ctrl-alt-Fx to other consoles if someone (want me to check or try something) For now, we boot on kernel 3.17.1-desktop-2.mga5 and it works nice.
CC: (none) => fri
@Morgan: does this have anything to do with the missing network / runlevel? I can't find any coherence to the threadtheme.
No idea, I am just an "advanced user"/reporter trying to bring information :) I can log and try things if i get specific instructions. Beware though this is a production machine although running cauldron, because neither mga4 nor fedora20 handle its suspend.
Whatever it was, that problem one of my copmuters had is gone with latest kernel 3.17.2-desktop-4.mga5 and other updates :)
Dear forumadmin, can you please move the comments from "Morgan Leijström" to the correct thread, because they have nothing to do with the "unknown runlevel/missing network" bug. After the last comment, people may think, that the bug is solved, but it isn't. The biggest problem is, that the bug occures if one tries to setup a computer (mostly used for server-services) with minimal software (like described in the first post). Clients with X-server and more rpm's installed does not have this problem. I did not found out, which rpm(s) is(are) needed to solve this problem.
Bugzilla is not a forum. Do you still have this problem with current cauldron? In your logs I see: shorewall[604]: /usr/lib64/ipset/ipset_bitmap_port.so: /usr/lib64/ipset/ipset_bitmap_port.so: undefined symbol: ipset_parse_tcp_udp_port This problem is supposed to be fixed now. --------------------------------------------------------------------------- Anyway, what you would want to know at first is: systemctl status multi-user.target (runlevels 2,3,4 are aliases of multi-user.target; runlevel 5 is an alias of graphical.target) On your system default.target is the same as multi-user.target as you said it is symlinked in /etc (I guess this determines your "runlevel" on boot up). If you don't see a problem there, the status of network.service network-up.service may be useful, and finally sshd.service but that's not where the problem is AFAICT.
CC: (none) => cjw
This problem is supposed to be fixed now. > --------------------------------------------------------------------------- > Anyway, what you would want to know at first is: systemctl status > multi-user.target (runlevels 2,3,4 are aliases of multi-user.target; > runlevel 5 is an alias of graphical.target) On your system default.target > is the same as multi-user.target as you said it is symlinked in /etc (I > guess this determines your "runlevel" on boot up). If you don't see a > problem there, the status of network.service network-up.service may be > useful, and finally sshd.service but that's not where the problem is AFAICT. Not fixed. After reboot: multi-user.target active network.service inactive (dead) network-up.service inactive (dead) sshd.service active (running)
***PUSH***
...and it's so easiliy to reproduce :-). I can't imagine, that solving this problem is such a big task, because if you run "init 3" manually after rebooting, everything is running fine. In my opinion, this entry is missing somewhere in the (self-generating?) start-script after the first update.
(In reply to Martin Schmitz from comment #22) > ...and it's so easiliy to reproduce :-). I can't imagine, that solving this > problem is such a big task, because if you run "init 3" manually after > rebooting, everything is running fine. You're probably right: For someone with plenty of free time and the needed knowledge it won't be hard. However, while Colin (the assignee) does have that knowledge, he does not have much time: he has a living to earn, a private life, is involved in at least one upstream project (if not more), has sysadmin tasks here, a team to lead, 58 more open bugs assigned to him etc. > > In my opinion, this entry is missing somewhere in the (self-generating?) > start-script after the first update. If you do happen to have time to inquire further, David's suggestion sounds like a good starting point (In reply to David Walser from comment #12) > Sounds to me like systemd isn't correctly setting the runlevel in utmp > anymore. Maybe systemd-update-utmp-runlevel.service or systemd-update-utmp.service is involved? I lack both the needed knowlegde and the needed time to look further
@ Martin Did you only hit this bug on the very first updates you did, and never again on later updates? When Meg reported he had hit it, it was about a week after the first Mageia 5 mass rebuild, in September. However, did it occur again when there were less updates at the same time? That might help to find the culprit (to me, with my less-than-basic understanding of how such things work, even initscripts and sysvinit-legacy-tools could be the cause)
For me, I hit the bug on every reboot ;) Didn't set up a new instance of mageia to check if I can run a "clean" system into the bug. (but comment #13 looks like that).
Ah, if it happens on _every_ reboot, and not only on the first reboot after updating for the first time, then I misunderstood :-/
Created attachment 6324 [details] Dead services after reboot
Created attachment 6325 [details] systemd-update-utmp-runlevel.status after reboot
Created attachment 6326 [details] systemd-update-utmp.status after reboot
sth. about time: I'm working fulltime in the university as admin, so I do not have much time either. I know, that you do you jobs for free and also in your free time, and I appreciate this. But the Mageia team asks for helping by bugreports. That's what I've done. Then it was written, it can't be this and that and the thread falls unsoluted in sleepmode. I can't even install a server, make an update, and then it is to be reinstalled again, because no other update does fix the network (same problem as you have: too less time). So I stopped trying it again. That was the state, when I reported this bug (installed it nearly 10 times new, always the same error). Now I live with an updateless server, that is rebootable. For me it's the best solution in the moment. I will setup a new Server, when 5 is finished. If the bug occures again, I will stay in this state, if not a bir security problem occures. But thanks at all for your work. I really like Mageia.
Could it be a duplicate to 15670 (which was fixed very recently)?
Keywords: (none) => NEEDINFO
Yep, by all the symptoms above it looks that way *** This bug has been marked as a duplicate of bug 15670 ***
Status: NEW => RESOLVEDCC: (none) => tmbResolution: (none) => DUPLICATE