Re: RelayChecker too slow?

[Justin Mason <jm...@jmason.org>]

John Rudd writes:
> Steven Manross wrote:
> >>> every message. JSA has an internal timeout of 20 sec which I cannot 
> >>> change without rebuilding the whole application. Is there a way to 
> >>> speed up RC?
> >>> I'm running the default setup with:
> >>> relaychecker_pass_auth  0
> >>> relaychecker_reduced_dns  0
> >> try setting that last one to 1.
> >>
> >> relaychecker_reduced_dns  1
> >>
> >>
> >> Let me know how that goes.
> >>
> > 
> > That took scan times from ~20 seconds down to 1-2 secs for me..
> > 
> > Steven 
> > 
> 
> I'm thinking about changing it so that one of the two optimizations in 
> reduced_dns is always true: use the "rdns=" part of the Untrusted Relay 
> pseudo-header directly.  Basically, trust SA and the MTA instead of 
> duplicating that effort.  Especially since that effort is made once for 
> each test.
> 
> Then it would just a matter of setting the score for 
> RELAY_CHECKER_BADDNS to 0 if you want to eliminate the other DNS check 
> that gets done.

hey --
I'd recommend it, fwiw...

--j.

Re: RelayChecker too slow?

[John Rudd <jr...@ucsc.edu>]

Justin Mason wrote:
> John Rudd writes:

>> I'm thinking about changing it so that one of the two optimizations in 
>> reduced_dns is always true: use the "rdns=" part of the Untrusted Relay 
>> pseudo-header directly.  Basically, trust SA and the MTA instead of 
>> duplicating that effort.  Especially since that effort is made once for 
>> each test.
>>
>> Then it would just a matter of setting the score for 
>> RELAY_CHECKER_BADDNS to 0 if you want to eliminate the other DNS check 
>> that gets done.
> 
> hey --
> I'd recommend it, fwiw...
> 

The next release of RelayChecker (0.4 or 1.0, not sure yet) now has this 
implemented.  So, right there, it will reduce the number of DNS checks, 
for the default configuration, from 5 to 1.  The "reduce_dns" option is 
removed, since the only thing you need to do now is set the score for 
RELAY_CHECKER_BADDNS to 0.

Also, RELAY_CHECKER_BADDNS now has a default score of 0.00 0.01 0.00 
0.01 ... so it should automatically obey if you've turned off network 
checks.

I'll probably release this new code next week some time.