[mythtv-users] Fedora, ceton, systemd, and mythbackend
Hika van den Hoven
hikavdh at gmail.com
Sun Jan 25 23:53:59 UTC 2015
Hoi Justin,
Monday, January 26, 2015, 12:32:06 AM, you wrote:
> I seem to run into a fun interplay between Fedora, my Ceton InfiniTV card,
> systemd, and my mythbackend (which handle two remote frontends).
> 1. The system boots.
> 2. The ceton driver loads, and for some reason systemd promptly renames
> this to network interface p5p1.
> 3. In the meantime, my actual ethernet interface (p5p2) is busy waiting on
> a response from my DHCP server.
> 4. Systemd loads mythbackend, which currently depends on
> network-online.target.
> 5. Because network interface p5p1 is up and running and has an IP address
> then YAY, as far as system is concerned the network is online. Mythbackend
> is a go to start.
> 6. Mythbackend starts, but doesn't actually listen on the not-yet-available
> ethernet connection to the outside world (p5p2).
> 7. At some point in the future, the DHCP server gets back to the backend
> and gives it a configuration for p5p2, meaning it can talk to the outside
> world. However it's too late for mythbackend to listen on this interface,
> and is not reachable from any of the remote frontends.
> This happens predictably and repeatably. I have no idea why systemd feels
> compelled to rename ctn0 to p5p1, or how to convince network manager that
> even though p5p1 looks and feels like a connection to the outside world,
> it's not. I already have mythbackend.service configured as starting
> "After=network-online.service" but clearly that's not enough.
> Google searches for "systemd network interface naming" and umpteen
> varieties of that only gives results talking about the wonderful new
> feature of systemd that will bring predictable network interface naming to
> Linux.
> First off, has anyone else seen this behavior?
> Secondly, does anyone have any suggestions on the right approach? Should I
> tweak systemd to not rename the interface? Should I try and get network
> manager to understand that p5p1 (or ctn0) is not actually a useful network
> interface? Should I just kludge it and put "sleep 10" before the
> mythbackend starts?
> Thanks,
> -jdm
A temporary workaround could be to put all processes needing network
on a 2? second wait-loop before starting up. Hopefully that would let
them get the right interface. Am I correct that systemd doesn't use
udev, but its own variant? Because that's where I would look.
Furthermore, I keep far from systemd for it's to inaccessible for
troubleshooting and I have heard and experienced to many things like
you name. I guess it needs another 5 years to mature.
Tot mails,
Hika mailto:hikavdh at gmail.com
"Zonder hoop kun je niet leven
Zonder leven is er geen hoop
Het eeuwige dilemma
Zeker als je hoop moet vernietigen om te kunnen overleven!"
De lerende Mens
More information about the mythtv-users
mailing list