Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.debian.user > #203947 > unrolled thread

odd resolver issue

Back to article view | Back to linux.debian.user

Contents

  odd resolver issue Andrea Borgia <andrea@borgia.bo.it> - 2019-01-04 11:10 +0100
    Re: odd resolver issue john doe <johndoe65534@mail.com> - 2019-01-04 11:20 +0100
      Re: odd resolver issue Andrea Borgia <andrea@borgia.bo.it> - 2019-01-04 12:50 +0100
    Re: odd resolver issue Martin <keine-eile@online.de> - 2019-01-04 11:50 +0100
      Re: odd resolver issue Andrea Borgia <andrea@borgia.bo.it> - 2019-01-04 12:50 +0100
        Re: odd resolver issue Andrea Borgia <andrea@borgia.bo.it> - 2019-01-04 13:10 +0100
        Re: odd resolver issue Martin <keine-eile@online.de> - 2019-01-04 13:20 +0100
          Re: odd resolver issue Andrea Borgia <andrea@borgia.bo.it> - 2019-01-04 13:40 +0100

#203947 — odd resolver issue

Hi.


Two systems on same LAN, let's call them A and B, with domain "localdomain".
Identical /etc/resolv.conf (generated by NetworkManager), including a 
"search localdomain" line.
Identical /etc/hosts with only localhost aliases in it.

 From A:
ping shortnameB (OK)
ping shortnameB.localdomain (OK)

 From B:
ping shortnameA (failure?!?!?)
ping shortnameA.localdomain (OK)

Both had DNSSEC turned off because of an annoying bug:
https://github.com/systemd/systemd/issues/10579

B is freshly booted and both are current "testing" setups.

What else should I check?


Thanks,
Andrea.

[toc] | [next] | [standalone]

#203948

On 1/4/2019 11:07 AM, Andrea Borgia wrote:
> Hi.
> 
> 
> Two systems on same LAN, let's call them A and B, with domain
> "localdomain".
> Identical /etc/resolv.conf (generated by NetworkManager), including a
> "search localdomain" line.
> Identical /etc/hosts with only localhost aliases in it.
> 
> From A:
> ping shortnameB (OK)
> ping shortnameB.localdomain (OK)
> 
> From B:
> ping shortnameA (failure?!?!?)
> ping shortnameA.localdomain (OK)
> 
> Both had DNSSEC turned off because of an annoying bug:
> https://github.com/systemd/systemd/issues/10579
> 
> B is freshly booted and both are current "testing" setups.
> 
> What else should I check?
> 

Are the dhcp server's options the same for both clients (identical lease)?

-- 
John Doe

[toc] | [prev] | [next] | [standalone]

#203954

Il 04/01/19 11:16, john doe ha scritto:


> Are the dhcp server's options the same for both clients (identical lease)?

Assuming OpenWRT isn't doing something fancy behind my back, the config 
entries look identical:

(this is global, so same lease)
config dhcp 'lan'
	option interface 'lan'
	option start '2'
	option limit '100'
	option leasetime '24h'
	option dhcpv6 'server'
	option ra 'server'
	option ra_management '1'
	option force '1'

(host A)
config host
	option name 'shortnameA'
	option ip 'x.y.z.4'
	option mac 'macA'

(host B)
config host
	option name 'shortnameB'
	option ip 'x.y.z.8'
	option mac 'wifimacB eth0macB'
(this is so that I can use either wifi or ethernet and still get the 
same address, which always worked fine since I never use both)

Given that /etc/resolv.conf is the same on both systems, as generated by 
NetworkManager, I would say the dhcp options are indeed just fine.


The fun thing is that this issue is new: it used to be that the first 
ping always failed (because of the DNSSEC bug I linked) but then it 
would work, regardless of which form, short or fqdn. Once I disabled 
DNSSEC, it would just work, always. Now, it makes no difference whether 
I use wifi or cable: ping to shortname fails, to fqdn works.

Of course, /etc/systemd/resolved.conf is the same: both have DNSSEC=off 
as the only difference with respect to the file as shipped.

[toc] | [prev] | [next] | [standalone]

#203952

Am 04.01.19 um 11:07 schrieb Andrea Borgia:
> Hi.
> 
> 
> Two systems on same LAN, let's call them A and B, with domain "localdomain".
> Identical /etc/resolv.conf (generated by NetworkManager), including a "search localdomain" line.

Add a 'domain localdomain'.
When using dhcp and this option is missing, have a look at /etc/dhcp/dhclient.conf.

> Identical /etc/hosts with only localhost aliases in it.
> 
> From A:
> ping shortnameB (OK)
> ping shortnameB.localdomain (OK)
> 
> From B:
> ping shortnameA (failure?!?!?)
> ping shortnameA.localdomain (OK)
> 
> Both had DNSSEC turned off because of an annoying bug:
> https://github.com/systemd/systemd/issues/10579
> 
> B is freshly booted and both are current "testing" setups.
> 
> What else should I check?
> 
> 
> Thanks,
> Andrea.
> 

[toc] | [prev] | [next] | [standalone]

#203956

Il 04/01/19 11:47, Martin ha scritto:


> Add a 'domain localdomain'.

To resolv.conf? Let me figure out first how to get N-M to do that, then 
there's always the question of the different behaviour.


> When using dhcp and this option is missing, have a look at /etc/dhcp/dhclient.conf.

The files are identical on both systems. Why is one of them consistently 
failing and the other one consistently working?



Off to see what N-M can be told to do...

[toc] | [prev] | [next] | [standalone]

#203959

Il 04/01/19 12:43, Andrea Borgia ha scritto:

> Il 04/01/19 11:47, Martin ha scritto:
>> Add a 'domain localdomain'.
> To resolv.conf? Let me figure out first how to get N-M to do that, then 
> there's always the question of the different behaviour.

Did a side-by-side comparison of the output of "nmcli con show", only 
two differences:
* IPv6 was disabled on the failing host: enabled, restarted the 
connection, no changes (as expected)
* the failing host has a 169.254.0.0 route, not relevant I'd say.


Puzzling...

[toc] | [prev] | [next] | [standalone]

#203961

Am 04.01.19 um 12:43 schrieb Andrea Borgia:
> Il 04/01/19 11:47, Martin ha scritto:
> 
> 
>> Add a 'domain localdomain'.
> 
> To resolv.conf? 

Yes.

> Let me figure out first how to get N-M to do that, then there's always the question of the different behaviour.
> 
> 
>> When using dhcp and this option is missing, have a look at /etc/dhcp/dhclient.conf.
> 
> The files are identical on both systems. Why is one of them consistently failing and the other one consistently working?

<sarcasm>
Because networkmanager knows everything better than anyone else.
</sacasm>
I have no domain record in resolv.conf as well, but my router is running a kind of dnsmasq. Fixed it there, once and for all.

> Off to see what N-M can be told to do...
> 

[toc] | [prev] | [next] | [standalone]

#203962

Il 04/01/19 13:19, Martin ha scritto:


> <sarcasm>
> Because networkmanager knows everything better than anyone else.
> </sacasm>

I hear you, I hear you :P


However, in this specific case it turned out that the problem was 
/etc/nsswitch.conf

working)
hosts:          files mdns4_minimal [NOTFOUND=return]  dns

failing)
hosts:          files mdns4_minimal [NOTFOUND=return] resolve 
[!UNAVAIL=return] dns



Thanks to everyone :)
Now I just have to figure out how that came to be and which is the 
default configuration.


Andrea.

[toc] | [prev] | [standalone]

Back to top | Article view | linux.debian.user

csiph-web