i'm seeing something similar where dnsmasq service is not waiting long enough. I have `listen-address=x.x.x.x` in my config and I see 


```
Oct 16 17:59:10 origin-master-1.localdomain systemd[1]: Started DNS caching server..
Oct 16 17:59:10 origin-master-1.localdomain dnsmasq[812]: dnsmasq: failed to create listening socket for x.x.x.x: Cannot assign requested address
Oct 16 17:59:10 origin-master-1.localdomain dnsmasq[812]: failed to create listening socket for x.x.x.x: Cannot assign requested address
Oct 16 17:59:10 origin-master-1.localdomain systemd[1]: dnsmasq.service: Main process exited, code=exited, status=2/INVALIDARGUMENT
```

if I restart dnsmasq after being booted for a while it works fine. I'll note I've seen this for
a while just haven't got around to reporting it. That means I've seen it on dnsmasq-2.76-3.fc26.x86_64
and the newer `2.76-5.fc26.x86_64`.

FYI: ubuntu bug for similar problem: https://bugs.launchpad.net/ubuntu/+source/dnsmasq/+bug/1531184

Petr, can we get you to look at this?

Hi,

I think this issue is better solved by using bind-dynamic instead of bind-interfaces. network-online.target is not used intentionally. If system is configured to use dnsmasq as DNS resolver of the system, network-online.target may not be active until dnsmasq is ready. It must not wait for network-online.target by default.

# dnsmasq.conf
interface=eth0
bind-dynamic

allows start of dnsmasq even when eth0 is still down. As soon as it is up, it will bind the interface and listen on it. I think this is exactly what is required in this case.

Is there reason why is bind-dynamic not suitable for your case? I would close the bug if it helps.

Thanks Petr. I'm going to work in an upstream issue [1] to see if we can use bind-dynamic.

[1] https://github.com/openshift/openshift-ansible/issues/5935

In the meantime, do you have any influence on dnsmasq upstream? It would be great if upstream could be modified to add a helpful message during boot. i.e. if I can't bind to eth0 when dnsmasq is started and bind-dynamic is not in the config then give a message to the user that says 'couldn't bind to address on eth0, maybe you would like to use bind-dynamic in dnsmasq config'.

(In reply to Petr Menšík from comment #3)
> Hi,
> 
> I think this issue is better solved by using bind-dynamic instead of
> bind-interfaces. network-online.target is not used intentionally. If system
> is configured to use dnsmasq as DNS resolver of the system,
> network-online.target may not be active until dnsmasq is ready. It must not
> wait for network-online.target by default.
> 
> # dnsmasq.conf
> interface=eth0
> bind-dynamic
> 
> allows start of dnsmasq even when eth0 is still down. As soon as it is up,
> it will bind the interface and listen on it. I think this is exactly what is
> required in this case.
> 
> Is there reason why is bind-dynamic not suitable for your case? I would
> close the bug if it helps.


Hi Petr

FANTASTIC!!  Using "bind-dynamic" works for me.

To answer your question: the reason I didn't use bind-dynamic is because I didn't know it existed!!  It's not mentioned anywhere in the comments of the sample dnsmasq.conf file.  I would suggest that it be added, as it's most useful in my case.

Many thanks.

Closing the bug, can be fixed by already supported configuration.

bind-dynamic is supported only on Linux, bind-interfaces on any other platform. I think default configuration and manual page are place for improvement. I guess code is not best place to document behaviour. Especially if failed bind to interface might be because incorrectly spelled interface name.

*** Bug 1984618 has been marked as a duplicate of this bug. ***