Bug 755682

Summary:	Can't login with any user (even root) making system 100% useless
Product:	[Fedora] Fedora	Reporter:	Terje Røsten <terje.rosten>
Component:	systemd	Assignee:	systemd-maint
Status:	CLOSED DUPLICATE	QA Contact:	Fedora Extras Quality Assurance <extras-qa>
Severity:	unspecified	Docs Contact:
Priority:	unspecified
Version:	16	CC:	johannbg, lpoetter, metherid, mschmidt, notting, plautrba, systemd-maint
Target Milestone:	---
Target Release:	---
Hardware:	Unspecified
OS:	Unspecified
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2011-12-14 02:51:13 UTC	Type:	---
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:

Description Terje Røsten 2011-11-21 18:53:04 UTC

Description of problem:

System configured with sssh with LDAP auth, home dirs on nfs.

Local fs:

 /
 /var
 /export
 
In login screen (runlevel 3) I only get:

[ 90.537402]: systemd-logind[863]: failed to start unit:Did not receive a reply. Possible causes include:...[cut]

then it just hangs forever.

In runlevel 5 I only get a completely empty screen except background image, after giving username and password.

What's really strange is going first to recovery mode (single to kernel command line),
and then:

service network start
init 3

there. And I can do a root text login and then do init 5.

And then X login works!

Seem to be a race condition or timing issue or something?

All updates installed. Even tried to install all testing updates, same result.

It's a bit hard to debug the broken system, not being able to login.

Any ideas?

Comment 1 Terje Røsten 2011-11-23 20:06:07 UTC

I finally managed to reproduce the problem with  one working virtual console letting me 
investigate the problem "live".

The systemd-logind was hanging, however when stopping the syslog-ng  service, 
systemd-logind started to work again. 

syslog-ng was set to log to remote log server in additional to local files.

I no idea of the root cause for this issue.

Comment 2 Michal Schmidt 2011-12-14 02:51:13 UTC

syslog-ng has some problems working with the socket from systemd.

*** This bug has been marked as a duplicate of bug 742624 ***

Comment 3 Terje Røsten 2011-12-14 12:47:17 UTC

Thanks for the hint Michal, issue seems identical.

However, I would still wish systemd was a bit more robust.

A system logger going wonky should not take down the whole system?

It's also very hard to bring the system back to working order.
The sys admin have to start in runlevel 1 or boot from CD (or similar) the get the system up and running.

Comment 4 Michal Schmidt 2011-12-14 13:28:47 UTC

(In reply to comment #3)
> However, I would still wish systemd was a bit more robust.
> 
> A system logger going wonky should not take down the whole system?

"wonky" in this case means the syslog received an open file descriptor for /dev/log, but never read from it. The system will be in a bad state when such a thing happens, whether using systemd or not. You can produce such a situation on a system with no systemd by sending SIGSTOP to the syslog process. You'll start seeing the same symptoms.

Comment 5 Terje Røsten 2011-12-14 15:35:28 UTC

Ok, thanks for your explanation.