www-builds mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Andrew Bayer <andrew.ba...@gmail.com>
Subject Re: Name resolution issue on ubuntu jenkins slaves?
Date Thu, 20 Aug 2015 17:16:53 GMT
Yup,
http://drewgraybeal.blogspot.com/2015/05/level-3-dns-hijacking-4222-and-others.html
-
level3 is hijacking NXDOMAIN now. I think Y!'s DHCP is setting those DNS
servers? Rajiv, can you comment?

A.

On Thu, Aug 20, 2015 at 10:32 AM, Andrew Bayer <andrew.bayer@gmail.com>
wrote:

> So the Yahoo! provided slaves (H*, ubuntu-*) do have Level3's DNS servers
> in /etc/resolv.conf - we don't actually control that directly, it's part of
> the system setup Y! uses, I think. I'll ping our contact at Y! about
> changing them to use more standard DNS.
>
> A.
>
> On Thu, Aug 20, 2015 at 6:15 AM, Gavin McDonald <gmcdonald@apache.org>
> wrote:
>
>>
>> > On 20 Aug 2015, at 4:03 am, David Nalley <david@gnsa.us> wrote:
>> >
>> > Whatever we are using for DNS on those build slaves is redirecting any
>> > NXDOMAIN to a Level3 search domain.
>> >
>> > Is this happening on the ubuntu-* slaves or on the dynamic build slaves?
>>
>> It was mentioned earlier :
>>
>> >  I am certainly seeing this on slaves ubuntu4
>> > through ubuntu6.
>>
>> I have seen passes and failures on the dynamic slaves too, so seems
>> inconsistent.
>>
>> Gav…
>>
>> >
>> > --David
>> >
>> > On Mon, Aug 10, 2015 at 8:59 AM, Keith W <keith.wall@gmail.com> wrote:
>> >> Hello Apache builds,
>> >>
>> >> We (Apache Qpid) have been seeing a test fail on the Ubuntu Jenkins
>> slaves
>> >> since August 9th.  The test verifies the behaviour of the code when the
>> >> user specifies a hostname that does not exist in DNS.  For this
>> purpose,
>> >> the test uses a random name 'hg3sgaaw4lgihjs' (without hierarchal part)
>> >> which is assumed not resolve.  This test is longstanding and has been
>> >> running on the slaves for many years without issue.
>> >>
>> >> Between August 9th and 10th, something appears to have changed on the
>> >> slaves, which is meaning that the lookup of the name is now returning
>> an
>> >> IP.  This is causing the Java test to fail. I've investigated by
>> >> introducing shell commands into the job, and can see evidence of the
>> same
>> >> problem at the UNIX level.  I am certainly seeing this on slaves
>> ubuntu4
>> >> through ubuntu6.
>> >>
>> >>
>> >> $ host hg3sgaaw4lgihjs
>> >> hg3sgaaw4lgihjs has address 198.105.244.11
>> >> hg3sgaaw4lgihjs has address 198.105.254.11
>> >> Host hg3sgaaw4lgihjs not found: 3(NXDOMAIN)
>> >>
>> >> $ host hg3sgaaw4lgihjs2
>> >> hg3sgaaw4lgihjs2 has address 198.105.244.11
>> >> hg3sgaaw4lgihjs2 has address 198.105.254.11
>> >> Host hg3sgaaw4lgihjs2 not found: 3(NXDOMAIN)
>> >>
>> >>
>> >> I considered changing the test to use a RFC-2606 Reserved Top Level DNS
>> >> Names  hg3sgaaw4lgihjs.invalid but I notice that it too is resolving to
>> >> 198.105.244.11 too.  The fact that an .invalid address is resolving
>> makes
>> >> me suspect there is an environmental problem at the root cause.
>> >>
>> >> host hg3sgaaw4lgihjs.invalid
>> >> hg3sgaaw4lgihjs.invalid has address 198.105.244.11
>> >> hg3sgaaw4lgihjs.invalid has address 198.105.254.11
>> >> Host hg3sgaaw4lgihjs.invalid not found: 3(NXDOMAIN)
>> >>
>> >> Is there a name resolution issue affecting these hosts?
>> >>
>> >> Example job affected by the issue:
>> >>
>> >>
>> https://builds.apache.org/view/M-R/view/Qpid/job/Qpid-Java-Java-Test-JDK1.8/
>> >>
>> >> Kind regards, Keith Wall.
>>
>>
>>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message