[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Configure resolv.conf Pods



Ok, to debug this, at the host level `dig @node-ip
kubernetes.default.svc.cluster.local` and `dig @node-ip www.google.es`
replace node-ip with the node's ip address, and then review the
journal for dnsmasq, `journalctl -lu dnsmasq`. If everything looks
fine there then I'd verify that the dnsIP is set correctly in
node-config.yaml

--
Scott

On Tue, Oct 4, 2016 at 11:35 AM, Fran Barrera <franbarrera6 gmail com> wrote:
> No, the problem isn't a conflict port because i've moved skydns to port
> 5353.
>
> Now, I have seen a bit more the problem, from a pod I can't resolver any
> openshift service and if I try a ping to www.google.es for example take ~ 10
> seconds. It's like dnsmasq is not forwarding to skydns? but my configuration
> is the same in both environments.
>
> Thanks,
> Fran.
>
>
>
>
> 2016-10-04 14:53 GMT+02:00 Scott Dodson <sdodson redhat com>:
>>
>> Hmm, I thought it was there in more recent versions too? The reason
>> that this is there is because we want 'fran.project' or 'frans' to use
>> the search path to resolve 'fransapp.project.svc.cluster.local' so
>> pods can reference the service without knowing the FQDN.
>>
>> Can you elaborate on how it's failing in the 1.0.8 environment? I
>> imagine the problem there is that there's probably a conflict on port
>> 53 on your masters which are also nodes? To work around this we've
>> moved skydns to port 8053, check master-config.yaml for
>> dnsConfig.bindAddress and see if it's binding to port 53 or 8053. If
>> you move it to 8053 and restart the master does it help?
>>
>> --
>> Scott
>>
>> On Tue, Oct 4, 2016 at 4:25 AM, Fran Barrera <franbarrera6 gmail com>
>> wrote:
>> > Hi Scott,
>> >
>> > I have configured dnsmasq as you commented and working good in my
>> > environment, but I have another environment in 1.0.8 version and the
>> > problem
>> > is the "option ndots:5" If i delete this option everything working good
>> > but
>> > I don't know why Openshift set this automatically in /etc/resolv.conf
>> >
>> > I see in other version that this options is dissapeared but in older
>> > version
>> > not. Is there any way to delete this option?
>> >
>> > Thanks,
>> > Fran
>> >
>> > 2016-10-03 14:43 GMT+02:00 Scott Dodson <sdodson redhat com>:
>> >>
>> >> Maybe I'm misunderstanding the question, but you'd configure dnsmasq
>> >> to forward 'cluster.local' to skydns and all other requests to your
>> >> desired nameservers, place a file in /etc/dnsmasq.d/ like this
>> >>
>> >> strict-order
>> >> no-resolv
>> >> domain-needed
>> >> server=/cluster.local/172.30.0.1
>> >> server=/30.172.in-addr.arpa/172.30.0.1
>> >> server=4.4.4.4
>> >> server=8.8.8.8
>> >>
>> >> Again, the installer does this for you automatically using a
>> >> NetworkManager dispatcher script to configure dnsmasq. You can see
>> >> what it does here
>> >>
>> >>
>> >> https://github.com/openshift/openshift-ansible/blob/master/roles/openshift_node_dnsmasq/files/networkmanager/99-origin-dns.sh
>> >>
>> >> On Mon, Oct 3, 2016 at 3:41 AM, Fran Barrera <franbarrera6 gmail com>
>> >> wrote:
>> >> > Hi Scott,
>> >> >
>> >> > If I set dnsmasq in the node-config.yaml. How I can forward cluster
>> >> > dns
>> >> > to
>> >> > skydns? Because Openshift services don't have any domain, right?
>> >> >
>> >> > Thanks,
>> >> > Fran
>> >> >
>> >> > 2016-09-30 14:56 GMT+02:00 Scott Dodson <sdodson redhat com>:
>> >> >>
>> >> >> Fran,
>> >> >>
>> >> >> If you set dnsIP in your node-config.yaml file it will make that IP
>> >> >> address the first nameserver in your pod's resolv.conf. I don't know
>> >> >> how to add timeout options to pod's resolv.conf, but hopefully you
>> >> >> should have timeout issues.
>> >> >>
>> >> >> BTW, the installer should be configuring dnsmasq on each node and
>> >> >> setting dnsIP to the local host's IP address starting with Origin
>> >> >> 1.2
>> >> >> and OCP 3.2. dnsmasq is then responsible for forwarding cluster dns
>> >> >> to
>> >> >> skydns and other requests to the host's default resolver.
>> >> >>
>> >> >> --
>> >> >> Scott
>> >> >>
>> >> >> On Fri, Sep 30, 2016 at 8:39 AM, Fran Barrera
>> >> >> <franbarrera6 gmail com>
>> >> >> wrote:
>> >> >> > Hi All,
>> >> >> >
>> >> >> > In my environment, the master is the DNS server (dnsmasq) and my
>> >> >> > problem
>> >> >> > is
>> >> >> > that I have very slow resolution. The resolv.conf of pods is:
>> >> >> >
>> >> >> > nameserver 172.30.0.1
>> >> >> > nameserver master_ip
>> >> >> >
>> >> >> > I need to change this conf adding first the master IP or add a
>> >> >> > options
>> >> >> > timeout because the slow is because try first with kubernetes dns
>> >> >> > and
>> >> >> > 10
>> >> >> > seconds later try with dns master. I've followed this guide
>> >> >> > https://access.redhat.com/solutions/2215521 but not change
>> >> >> > anything.
>> >> >> >
>> >> >> > Any helps?
>> >> >> >
>> >> >> >
>> >> >> >
>> >> >> > _______________________________________________
>> >> >> > users mailing list
>> >> >> > users lists openshift redhat com
>> >> >> > http://lists.openshift.redhat.com/openshiftmm/listinfo/users
>> >> >> >
>> >> >
>> >> >
>> >
>> >
>
>



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]