On Mar 17, 2014, at 11:38, "Adrian Schröter"
wrote: On Montag, 17. März 2014, 11:31:04 wrote Bahi, David:
had to play some device name games to get the DHCP working on 1st boot (eth0 -> ens2f0 for ifcfg- and a udev rule)
are you sure you had this problem with .98? It is supposed to be fixed there, enforcing udev to use eth0 always.
Yes because it was assigning the wrong interface eth 0
What exactly was the case directly after boot:
* You had an interface called eth0 without network?
Yes - this - had eth0 without network and other three interfaces had 'rename#' for their ifname. eth0 was ifname was associated with 2nd (or on another try the 3rd) network interface - but only first was wired.
* You had an interface ens2f0 without network?
Not this. But after xzcat and before first boot of OBS my playing with udev and ifcfg file in /dev/sda1 allowed ens2f0 to be assigned to the correct interface and have DHPC and DNS working for the first boot configuration.
* something else?
boot's and tells me all is well at prompt.
browse to https://<fqdn>/ and see:
Web application could not be started An error occurred while starting up the preloader. It exited before signalling successful startup back to Phusion Passenger. Please read this article for more information about this problem. Raw process output:
most likely a follow up problem due to invalid web/rails configuration due to to invalid hostname
Unlikely unless there is a dependency on the eth 0 ifname
There is a dependency that a work DHCPD and DNS server was found before the obs*setup scripts did run on boot time.
And that is working - I do not see the bad message about network not working before the prompt. Networking is good but the ifname is not eth0 and the above error message is seen. Haven't tried with ifname as eth0 (just messing with udev rules to get it bound to the first interface - in that pre-deploy play) So think it could be a deprecate issue.