Hello,

guardian parameter is not present in the default configuration file.
I plan to check the performance but the first step is too ensure all is working the same then if we improve the performance it is a bonus.

I will get back to you with more details.

Regards,
Francois

On 2012-08-31 14:05, Ruben d'Arco wrote:
Hi,

On Fri, Aug 31, 2012 at 01:30:39PM +0200, xbgmsharp wrote:
Hello,

Thanks for the reply.
I use PowerDNS's SVN version as TinyDNS backend is not ship by
default in the package.
So I believe, I have this patch apply,
http://wiki.powerdns.com/trac/changeset/2622

Yep!

I will apply the patch from github.
According to my first test using dnsperf, PowerDNS with TinyDNS
backend seen faster than TinyDNS himself.
The biggest concern is stability to avoid PowerDNS to crash due to
weird data or too many data.
The file contains 26643 entries and errors.

PowerDNS can run with a guardian, so it would never crash and burn
without launching again.
This can be set using the guardian=yes option in the config file (i'm
not sure if it's a default).

The patch mentioned on github should avoid errors whenever there are
errors in the data.cdb file.
All the times this happend, we could trace it down to an error in the
data file, which results in wrong
data in the data.cdb file. Lot's of those errors are mentioned here:

http://homepage.ntlworld.com/jonathan.deboynepollard/FGA/djbdns-problems.html#tinydns-data-semantic-error
But, as you're a tinydns user, you'll probably already know about these :-)

It use the default configuration from PowerDNS, I am sure it could
be improve by tunning PowerDNS.

The performance improvement probably comes from caching that PowerDNS has.
http://doc.powerdns.com/performance-settings.html might be nice to
read to gain some more understanding
of this.

The system is Ubuntu 12.04 compile with gcc 4.6.3 on x86_64.
The hardware is a one Intel(R) Xeon(R) CPU X3330 @2.66GHz with 4
Cores and 4096M System RAM.
It is not the production system but should give us a nice idea.
I need to go deeper in the testing, but so far so good.

PowerdNS with tinyDNS backend.
Statistics:

  Queries sent:         26608
  Queries completed:    15416 (57.94%)
  Queries lost:         11192 (42.06%)

  Response codes:       NOERROR 15415 (99.99%), NXDOMAIN 1 (0.01%)
  Average packet size:  request 45, response 61
  Run time (s):         565.866733
  Queries per second:   27.243164

  Average Latency (s):  0.035546 (min 0.033069, max 0.056192)
  Latency StdDev (s):   0.001457

TinyDNS
Statistics:

  Queries sent:         26608
  Queries completed:    11835 (44.48%)
  Queries lost:         14773 (55.52%)

  Response codes:       NOERROR 11834 (99.99%), NXDOMAIN 1 (0.01%)
  Average packet size:  request 45, response 129
  Run time (s):         745.820314
  Queries per second:   15.868433

  Average Latency (s):  0.049470 (min 0.041367, max 0.078622)
  Latency StdDev (s):   0.004388

Regards,
Francois


Regards,
        Ruben

On 2012-08-31 10:18, Ruben d'Arco wrote:
>Hi Francois,
>
>The tinydnsbackend is marked experimental because 3.1 is the first
>release that has the backend.
>PowerDNS needs some 'out in the field' validation that the backend is
>working correctly.
>Please provide feedback, as we could then remove the experimental
>flag!
>
>We know one user who uses it with a data.cdb file > 70M. That user
>also has identified some issues, one
>important issue is fixed in
>http://wiki.powerdns.com/trac/changeset/2622
>You can either apply the patch yourself, or take PowerDNS's SVN
>version to have this resolved.
>This patch is strongly recommended!
>
>The user also reported that PowerDNS handles errors in the data.cdb
>file different from tinydns.
>I've added an configuration option to allow the behaviour to be
>configurable, the patch for that can be found here:
>

>https://github.com/cyclops1982/powerdns/compare/master...tinydns2.diff
>This patch is only useful if your data.cdb file is not a 100%
>correct. Something that can be resolved by
>making sure your data file is correct before running tinydns-data.
>
>The scenario you're planning (move from dbjdns to powerdns and then
>change backend) is a scenario we've heard before
>and partially why the backend was created - the master mode that the
>backend provides should help you with this.
>
>Kind regards,
>        Ruben d'Arco
>
>
>
>On Fri, Aug 31, 2012 at 09:53:58AM +0200, xbgmsharp wrote:
>>Hello,
>>
>>According to the documentation
>>(http://doc.powerdns.com/tinydnsbackend.html), the TinyDNS backend
>>is marked as experimental!
>>It this statement still true?
>>
>>We are currently using tinydns and djbdns in all our infrastructure.
>>We are thinking in moving to Powerdns to support new feature.
>>However before doing a big move, we would rater make the move in
>>multiple step.
>>The first step would be to use Powerdns with TinyDNS backend.
>>The second step would be to move from TinyDNS backend to a database
>>backend.
>>This second step require a rewrite of our DNS tools and it will take
>>longer.
>>
>>Our data.cdb is file is around 18M.
>>
>>Do you think it is safe to use Powerdns with TinyDNS backend in
>>production?
>>
>>Regards,
>>Francois
>>
>>_______________________________________________
>>Pdns-users mailing list
>>[email protected]
>>http://mailman.powerdns.com/mailman/listinfo/pdns-users
>_______________________________________________
>Pdns-users mailing list
>[email protected]
>http://mailman.powerdns.com/mailman/listinfo/pdns-users

_______________________________________________
Pdns-users mailing list
[email protected]
http://mailman.powerdns.com/mailman/listinfo/pdns-users
_______________________________________________
Pdns-users mailing list
[email protected]
http://mailman.powerdns.com/mailman/listinfo/pdns-users

_______________________________________________
Pdns-users mailing list
[email protected]
http://mailman.powerdns.com/mailman/listinfo/pdns-users

Reply via email to