[Dnsmasq-discuss] Stats improvement

Dave Taht dave.taht at gmail.com
Fri Mar 28 16:36:12 UTC 2014


On Fri, Mar 28, 2014 at 9:35 AM, Dave Taht <dave.taht at gmail.com> wrote:
> On Thu, Mar 27, 2014 at 1:57 PM, Simon Kelley <simon at thekelleys.org.uk> wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> On 26/03/14 05:12, Olivier Mauras wrote:
>>> Yes it should definitely be TXT records. Sounds really good to me.
>>>
>>> for upstream servers, why not having upstream.bind return total
>>> queries forwarded + list of upstream servers that can then be check
>>> individually with: dig TXT <server1>.upstream.bind
>>
>> OK. Code in git now.
>>
>> Try eg
>>
>> dig +short chaos txt misses.bind
>>
>> and
>>
>> dig +short chaos txt servers.bind
>
> I haven't looked at the code yet, but my take on this was originally that
> I'd wanted a single dns reply for all the statistics (can live
> without), 64 bit counters, and not only tracking misses but the
> latency cost of a miss. Then there are misses that are never

latency cost of a fill, I meant. This would end up giving stats that
are comparable to namebench.


> resolved... and (even harder), things like a 5 minute rolling
> average... and a breakdown by query type (AAAA, A, DS, etc)
>
> gettimeofday() is very efficient on modern linuxes (it gets the time from a
> shared page cache without even making a syscall), but I can see that
> you might want to be able to compile that out.
>
> And I wasn't expecting you to get to it in this release! Just basic stats is
> great....
>
>
>>
>>
>> Cheers,
>>
>> Simon.
>>
>>>
>>>
>>> Olivier
>>>
>>> On Tue, 2014-03-25 at 21:37 +0000, Simon Kelley wrote:
>>>> My feeling it that this should use TXT records, then they can be
>>>> extracted into any application using eg
>>>>
>>>> dig +short chaos txt stats.bind
>>>>
>>>> or the equivalent, which outputs the string to stdout.
>>>>
>>>> The current code has,
>>>>
>>>>
>>>> cache-size, insertions which evicted unexpired cache entries,
>>>> queries forwarded, queries answered locally, queries for auth
>>>> zones.
>>>>
>>>>
>>>> then for each upstream server,
>>>>
>>>> queries sent, and queries retried or failed.
>>>>
>>>> We could have
>>>>
>>>> size.bind and evictions.bind for cache size and evictions
>>>> local.bind, upstream.bind and auth.bind for queries.
>>>>
>>>> But how to handle the variable number of upstream servers. Just
>>>> free text?
>>>>
>>>>
>>>>
>>>> Simon.
>>>>
>>>> Enigmail On 24/03/14 22:53, Olivier Mauras wrote:
>>>>>
>>>>>
>>>>> On Mon, 2014-03-24 at 21:55 +0000, Simon Kelley wrote:
>>>>>> On 24/03/14 11:25, Olivier Mauras wrote:
>>>>>>>
>>>>>>>
>>>>>>> Hello,
>>>>>>>
>>>>>>> I was wondering what would be the effort, and if there'd
>>>>>>> actually be any interest for some dnsmasq statistics
>>>>>>> improvements. (Yes i'm splitting dicussions ^^) For
>>>>>>> monitoring/graph purposes, actual dnsmasq stats are a bit
>>>>>>> difficult to use and completely unusable if using
>>>>>>> "log_queries" as it takes too long to retrieve them inside
>>>>>>> logs.
>>>>>>>
>>>>>>> I'd love to see a stats "interface" that would output
>>>>>>> total_queries, cache_hits, cache_misses, memory used by
>>>>>>> cache, etc....
>>>>>>>
>>>>>>> Thanks, Olivier
>>>>>>>
>>>>>>
>>>>>> There's an idea to make this available as a DNS query, in
>>>>>> the same way that
>>>>>>
>>>>>>
>>>>>> dig chaos txt version.bind
>>>>>>
>>>>>> returns the version number.
>>>>>>
>>>>>> Comments?
>>>>>>
>>>>>> Cheers,
>>>>>>
>>>>>> Simon.
>>>>>>
>>>>>>
>>>>>
>>>>> That would be a really good way of doing it!
>>>>>
>>>>
>>>
>>
>> -----BEGIN PGP SIGNATURE-----
>> Version: GnuPG v1.4.11 (GNU/Linux)
>> Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/
>>
>> iEYEARECAAYFAlM0kMsACgkQKPyGmiibgrcBGQCgh70G5hX6LqTdO+2z0ZEg0PYe
>> 25EAnRvZ79kpqBVa4g9dGz7XXlKCOTR5
>> =85b/
>> -----END PGP SIGNATURE-----
>
>
>
> --
> Dave Täht
>
> Fixing bufferbloat with cerowrt: http://www.teklibre.com/cerowrt/subscribe.html



-- 
Dave Täht

Fixing bufferbloat with cerowrt: http://www.teklibre.com/cerowrt/subscribe.html



More information about the Dnsmasq-discuss mailing list