<div dir="ltr"><div>Oops, I think I missed the attachments. Here they go.<br></div>Greg<br></div><div class="gmail_extra"><br><br><div class="gmail_quote">2013/5/14 Gergely Bácskai <span dir="ltr"><<a href="mailto:mor.mango@gmail.com" target="_blank">mor.mango@gmail.com</a>></span><br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><div><div><div><div><div><div><div><div>Yep, all queries were identical.<br></div>I've read some documentation of the queryperf, it's a DNS-server performance measuring tool, shipped with BIND (as I know).<br>
</div>From this info I think the "query/sec" was low because of my virtualbox, but I'm not sure about this..<br></div><div><br></div>...but
this should not affect the count of success/failure. As I know,
queryperf does not respond to the truncated responses. I'm not sure if
it counts these as "successful responses", I'll look after it.<br></div>I
have set the logging in my BIND server, but the only log related to RRL
I could find is in the "/var/log/named/bind-queries.log"<br><br></div><div>I've made another try with 10 queries, and the RRL settings were: slip=2, resposnses-per-second=5, window=5, I've attached the results on 2 images: one is the queryperf result, and the other one is the bind-queries.log part of my action. Can somebody tell me from the attached bind log, how many queries were successful of the 10, so I can correlate it with the queryperf result?<br>
<br></div></div>(Or maybe could anybody suggest a method that I can use to get the "expected" results of the mentioned RRL-settings?)<br><br></div>Thanks,<br></div>Greg</div><div class="HOEnZb"><div class="h5">
<div class="gmail_extra"><br><br><div class="gmail_quote">
2013/5/10 Vernon Schryver <span dir="ltr"><<a href="mailto:vjs@rhyolite.com" target="_blank">vjs@rhyolite.com</a>></span><br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
> From: =?ISO-8859-1?Q?Gergely_B=E1cskai?= <<a href="mailto:mor.mango@gmail.com" target="_blank">mor.mango@gmail.com</a>><br>
<div><br>
> I set the RRL's window size to 5, responses-per-second to 5, and just for<br>
> sure, slip to 2.<br>
><br>
> I have sent 1000 queries to the server with the queryperf script.<br>
><br>
> The results are:<br>
> 1000 queries, runtime: 131,8 sec, -->7.58 queries/sec<br>
> ~50% success ->3.8 successful queries/sec<br>
><br>
> My two questions are:<br>
> -how does the window size=5 affect in this example? (I've tried it with<br>
> window=1, and got almost the same results, 57% success...)<br>
<br>
</div>There is this text for the ARM:<br>
<br>
Rate limiting uses a "credit" or "token bucket" scheme. Each<br>
identical response has a conceptual account that is given<br>
responses-per-second, errors-per-second, and nxdomains-per-second<br>
credits every second. A DNS request triggering some desired<br>
response debits the account by one. Responses are not sent while<br>
the account is negative. The account cannot become more positive<br>
than the per-second limit or more negative than window times the<br>
per-second limit. A DNS client that sends requests that are not<br>
answered can be penalized for up to window seconds (default 15).<br>
<div><br>
> -Shouldn't it act like:<br>
> "I have 1000 queries, 131.8 sec runtime, so because of the<br>
> responses-per-sec=5, every second it should be at first 5 successful<br>
> queries -->5x131.8=659 successful queries at all minimum.<br>
> But because of the slip=2, every second query of the rest 1000-659=341<br>
> (about 170 queries) should be also successful.<br>
> So at all, it should be 170+659=859 successful queries out of 1000?"<br>
> ...And I don't even know how to count the window=5 for this....<br>
<br>
</div>Were all of the queries identical?<br>
<br>
Does queryperf count "slipped" or truncated (TC=1) responses as<br>
"successful"? Does querfyperf react to truncated responses by<br>
retrying with TCP?<br>
I suspect queryperf is doing at least some retrying and delaying<br>
for responses before sending more queries because 8 queries/second<br>
is too slow for hardware made in this century.<br>
<br>
A test that sends 1000 queries in one second should receive 5 normal<br>
response followed by 497 or 498 "slipped" or truncated (TC=1)<br>
responses.<br>
<br>
The only good way I can see to answer such questions is to configure<br>
BIND9 logging for the "queries" and "rate-limit" categories with a<br>
debug level of at least 1, and look. That will tell when BIND received<br>
each query and what did.<br>
<br>
<br>
Vernon Schryver <a href="mailto:vjs@rhyolite.com" target="_blank">vjs@rhyolite.com</a><br>
_______________________________________________<br>
ratelimits mailing list<br>
<a href="mailto:ratelimits@lists.redbarn.org" target="_blank">ratelimits@lists.redbarn.org</a><br>
<a href="http://lists.redbarn.org/mailman/listinfo/ratelimits" target="_blank">http://lists.redbarn.org/mailman/listinfo/ratelimits</a><br>
</blockquote></div><br></div>
</div></div></blockquote></div><br></div>