We are using fedora-readonly.service in a netboot environment where we boot via NFS.
With the latest upgrade of NetworkManager to 1:0.9.7.997-2.fc18 the setup does not start correctly anymore.
With NetworkManager.x86_64 1:0.9.7.0-12.git20121004.fc18 everything worked fine.
I am trying to find out if it's one of the fedora specific problems or the general update to 0.9.8 which causes the problem. Sadly I did not find a way to debug, yet as the system simply freezes during boot, with the latest version of NetworkManager.
Yeah, that's not easy to debug things when the system doesn't boot.
But, you may look at http://freedesktop.org/wiki/Software/systemd/Debugging that could help finding out what's going on, capture logs, etc.
It may be that the Dracut scripts which move the DHCP lease to the real system after switchroot don't move them to the new location that NM 0.9.8 looks for them in. Also, NetworkManager doesn't look in the old location for leasefiles either.
Posted a patch to the NM mailing list that may solve this, and also did this scratch build. Are you able to install this scratch build and test whether it fixes the issue?
I'm handling the case for Marcus. I installed NetworkManager 0.9.8.0-1.1, but the freeze remains. I managed to boot single user mode and start NetworkManager manually. I will try to get some debugging output to provide more information.
Created attachment 710595 [details]
NetworkManager --no-daemon --log-level=DEBUG
After starting NetworkManager, system freezes
Yes, NM is taking the interface down, which it shouldn't be doing. I'll post a new scratch build with more debugging if you would be so kind as to repeat the testing. Thanks!
Actually, here's one issue:
NetworkManager: <warn> Failed to copy leasefile /var/lib/dhclient/dhclient-6af563d3-2c30-4e67-ab90-3a353f852b09-eth0.lease to /var/lib/NetworkManager/dhclient-6af563d3-2c30-4e67-ab90-3a353f852b09-eth0.lease: (21) Error opening file '/var/lib/NetworkManager/dhclient-6af563d3-2c30-4e67-ab90-3a353f852b09-eth0.lease': Read-only file system
is /var/lib/ really read-only at this point? That seems odd; /var/lib/ is usually supposed to be read/write like /var/run, I think. In any case, I'll see if I can modify the patch to just use the old location instead of copying it.
In our current stable netboot environment, /var/lib is r/o. Some of the subdirectories are replaced with tmpfs and made r/w, when booting properly. Inclunding /var/lib/dhclient, but not yet /var/lib/NetworkManager.
I could not yet figure out from single user mode if during normal boot, these tmpfs mounts are completed when NetworkManager wants to do the copy job. And I am still searching the code that triggers the mounts, if I find it I could test with an additional tmpfs mount for /var/lib/NetworkManager.
Added /var/lib/NetworkManager to rwtab, system boots nicely, no locking anymore.
Proposed patch for review upstream in associated GNOME bug. That patch should fall back correctly if it cannot copy the leasefile.
@Bill: could we please add:
Added on git master.
network-manager-applet-0.9.8.1-1.git20130327.fc18,NetworkManager-0.9.8.1-1.git20130327.fc18 has been submitted as an update for Fedora 18.
Package network-manager-applet-0.9.8.1-1.git20130327.fc18, NetworkManager-0.9.8.1-1.git20130327.fc18:
* should fix your issue,
* was pushed to the Fedora 18 testing repository,
* should be available at your local mirror within two days.
Update it with:
# su -c 'yum update --enablerepo=updates-testing network-manager-applet-0.9.8.1-1.git20130327.fc18 NetworkManager-0.9.8.1-1.git20130327.fc18'
as soon as you are able to, then reboot.
Please go to the following url:
then log in and leave karma (feedback).
network-manager-applet-0.9.8.1-1.git20130327.fc18, NetworkManager-0.9.8.1-1.git20130327.fc18 has been pushed to the Fedora 18 stable repository. If problems still persist, please make note of it in this bug report.