Bug 1332442

Summary: Kickstart network installation over VLAN hangs at reboot
Product: Red Hat Enterprise Linux 7 Reporter: nico.martin
Component: NetworkManagerAssignee: Rashid Khan <rkhan>
Status: CLOSED DUPLICATE QA Contact: Desktop QE <desktop-qa-list>
Severity: high Docs Contact:
Priority: high    
Version: 7.2CC: aloughla, atragler, bgalvani, lrintel, mleitner, nico.martin, rkhan, systemd-maint-list, thaller
Target Milestone: rc   
Target Release: ---   
Hardware: x86_64   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2016-07-04 09:57:52 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Attachments:
Description Flags
Result of mount command in anaconda
none
Anaconda log - journal.log
none
Anaconda log - anaconda.log
none
Anaconda log - ifcfg.log
none
Anaconda log - storage.log
none
Anaconda log - syslog none

Description nico.martin 2016-05-03 08:23:01 UTC 
Description of problem:
Following setup is used:
- kickstart has a reboot command
- kickstart is fetched on a dhcp server, over a VLAN (vlan id 8 in our case)
- installation is done through the network (NFS server reached through vlan id 8)
- installation is stuck at the end, the machine does not reboot.

Version-Release number of selected component (if applicable):
anaconda-21.48.22.56-1.el7

How reproducible:
Systematic

Steps to Reproduce:
1. Write a kickstart file with the command reboot in it. The repo used for installation has to be located on a NFS server that is reached through a VLAN.
2. Set up dhcp server that will serve the kickstart over a VLAN
3. On target machine, specify anaconda boot parameters "vlan=eth2.8:eth2 ks"
4. Let the installation run
5. Installation will hang at reboot. Control + Alt + FX does not work, Control + Alt + Del does not work.

Actual results:
Installation is stuck at the end.

Expected results:
Machine reboots to newly installed system.

Additional info:
When the command "systemctl --no-wall reboot" is issued by anaconda, NetworkManager will bring down the parent interface of the VLAN (eth2 in our case).
Since we have the NFS mount through the VLAN interface, system is stuck.
Comment 2 Brian Lane 2016-05-06 00:10:02 UTC 
Please attach the logs from /tmp/*log as individual text/plain attachments.

Also include the output of mount, this sounds similar to the iscsi problems where dracut didn't know it was a network mount and wasn't umounting in the right order.
Comment 3 nico.martin 2016-05-06 09:45:57 UTC 
Created attachment 1154533 [details]
Result of mount command in anaconda
Comment 4 nico.martin 2016-05-06 09:47:08 UTC 
Created attachment 1154534 [details]
Anaconda log - journal.log
Comment 5 nico.martin 2016-05-06 09:47:36 UTC 
Created attachment 1154535 [details]
Anaconda log - anaconda.log
Comment 6 nico.martin 2016-05-06 09:48:02 UTC 
Created attachment 1154536 [details]
Anaconda log - ifcfg.log
Comment 7 nico.martin 2016-05-06 09:48:30 UTC 
Created attachment 1154538 [details]
Anaconda log - storage.log
Comment 8 nico.martin 2016-05-06 09:49:56 UTC 
Created attachment 1154540 [details]
Anaconda log - syslog
Comment 9 nico.martin 2016-05-06 09:51:31 UTC 
There you go !
Let me know if you need more logs.
Comment 10 Brian Lane 2016-05-09 18:46:05 UTC 
systemd should be unmounting the nfs mounts before shutting down the network.
Comment 11 Lukáš Nykrýn 2016-05-10 10:10:56 UTC 
Looks like the issues with missing After=dbus.service inside NetworkManager.service. If I am wrong feel free to reassign this back.
Comment 12 nico.martin 2016-05-11 09:34:10 UTC 
We found a workaround in the meantime; if we assign an IP address to the parent interface (eth2) using the network command in kickstart, then the machine reboots after installation.

I also tested Lukas' suggestion and manually added "After=dbus.service" to NetworkManager.service during installation and the machine reboots correctly after installation.
Comment 13 Beniamino Galvani 2016-07-04 09:40:35 UTC 
(In reply to nico.martin from comment #12)
> I also tested Lukas' suggestion and manually added "After=dbus.service" to
> NetworkManager.service during installation and the machine reboots correctly
> after installation.

Does this work even without the workaround of assigning an IP address to the parent interface? If so, probably we can close this as duplicate of bug 1311988
Comment 14 nico.martin 2016-07-04 09:50:10 UTC 
Yes, manually adding "After=dbus.service" to NetworkManager.service makes it work without any other modification/workaround.
Comment 15 Beniamino Galvani 2016-07-04 09:57:52 UTC 
(In reply to nico.martin from comment #14)
> Yes, manually adding "After=dbus.service" to NetworkManager.service makes it
> work without any other modification/workaround.

Thanks, this has been fixed in NM 1.2.

*** This bug has been marked as a duplicate of bug 1311988 ***