Mailinglist Archive: opensuse-bugs (4233 mails)

< Previous Next >
[Bug 1020301] New: dbus service restart in TW 20170112 update crashed sytemd-logind
  • From: bugzilla_noreply@xxxxxxxxxx
  • Date: Tue, 17 Jan 2017 08:53:17 +0000
  • Message-id: <bug-1020301-21960@http.bugzilla.opensuse.org/>
http://bugzilla.opensuse.org/show_bug.cgi?id=1020301


Bug ID: 1020301
Summary: dbus service restart in TW 20170112 update crashed
sytemd-logind
Classification: openSUSE
Product: openSUSE Tumbleweed
Version: Current
Hardware: x86-64
OS: SUSE Other
Status: NEW
Severity: Major
Priority: P5 - None
Component: Upgrade Problems
Assignee: bnc-team-screening@xxxxxxxxxxxxxxxxxxxxxx
Reporter: martin.wilck@xxxxxxxx
QA Contact: jsrain@xxxxxxxx
Found By: ---
Blocker: ---

Created attachment 710346
--> http://bugzilla.opensuse.org/attachment.cgi?id=710346&action=edit
journalctl output during update

This is a follow-up on the disussion on opensuse-factory ML subject "CAREFUL:
New Tumbleweed snapshot 20170112 released"

The problem occured while I was running "zypper dup --no-allow-vendor-change"
in screen session in a gnome-terminal window under GNOME. Admittedly, this was
risky business.

On my system, "systemctl daemon-reload" happened 9 times during
the update which eventually crashed. But that alone wasn't fatal. The
fatal problem was the restart of the dbus service, which caused various other
services to be stopped and restarted as well, including systemd-logind. systemd
and other services failed to create dbus connections. In the wake of these
events, the gdm session and X crashed. systemd was again reloaded while
services were restarted. systemd started to emit the error message "Looping too
fast. Throttling execution a little" - probably while it was trying to
restart systemd-logind. The restart of the systemd-logind service
eventually failed, which explains why I wasn't able to log in on the
console to see what went wrong.

The logs show that rpm continued updating packages in spite of these
errors. Even the initrd seems to have been rebuilt. But no btrfs "post"
transaction snapshot has been created, so the zypper transaction didn't
fully succeed.

AFAICS, the problem was caused by the restart of the dbus service in the
%postuninstall section of the dbus-1 package. On my system,
DISABLE_RESTART_ON_UPDATE in /etc/sysconfig/services has the default value
"no", so in a way this behaved as I configured it. But maybe the dbus service
should be an exception from this rule, or should be controlled by a separate
option, e.g. "ENABLE_DANGEROUS_RESTART_ON_UPDATE='yes'".

A log excerpt is attached. At the end it shows my attempts to determine the
status of the system using sysrq, as no login was possible any more.

I recovered from this problem by booting from a btrfs snapshot and doing the
upgrade again in a text console.

--
You are receiving this mail because:
You are on the CC list for the bug.
< Previous Next >