On Sat, 4 Apr 2020 at 15:23, Tom Horsley horsley1953@gmail.com wrote:
I'm using network, not NetworkManager. I've got a bridge set up with my (one and only) ethernet point connected to it. I'm running fedora 31.
Randomly, when I reboot the system to get a new kernel
or something, dhcp will fail to assign an IP address to the bridge. I'll reboot again and it will be OK (I have also been able to do an ifdown/ifup to get an IP addr).
A bridge doesn't require an IP address. At my former work, we were only allowed to connect one device to each ethernet drop. Even so there were some problems tracking down unauthorized devices doing bad things on the network.
I once found a web page that said STP (spanning tree protocol) slowed things way down, and with my config I can't possibly have any loops, so I disable STP. That did indeed make network startup far more reliable, but it still sometimes fails.
Does anyone else have this problem? Is there some magic I can perform to make it reliable? I don't have this problem when configured with a static IP, but they were redoing a lot of network infrastructure and recommended everyone switch to dhcp.
At my work, this was done because there were more devices than available addresses. If you rebooted, you risked not being able to get an IP. We negotiated for a small block of fixed addresses for the servers in my group.
It also seemed very reliable before I created the bridge (but I need the bridge to run virtual machines).
Working from home, I've now become paranoid about installing updates that need a reboot for fear I won't be able to get to the machine again because the network never comes back up.
Maybe I should install a script in rc.local that checks to see if an IP address was defined, and reboots again if not :-).
If you can do a script you could just try getting dhcp to assign an address in a loop until it works.