Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
Micah
@micah_gitlab
status: NOERROR and I get a list of root nameservers
Micah
@micah_gitlab
how can I get kdig to do the same thing as+trace
Micah
@micah_gitlab
the problem is, i get a SERVFAIL in response to a lookup, then knot seems to cache it for longer than it should
SERVFAIL is covered by §7.1 of RFC2308:
In either case a resolver MAY cache a server failure response. If it does so it MUST NOT cache it for longer than five (5) minutes, and it MUST be cached against the specific query tuple <query name, type, class, server IP address>.
Vladimír Čunát
@vcunat
kresd does not cache SERVFAILs at all. That's actually a feature I want to implement (some day).
There is some caching of servers that repeatedly don't reply, so that can manifest as SERVFAIL that's returned immediately "from cache".
But by default it's very short now IIRC, just 1s or 5s.
beckhamaaa
@beckhamaaa
ok, i see.
when i insert the rr to the cache by lua script , count is 0.1 billion, the insert speed is only serveral hundred, how can i optimum it ? i thank it's the performmance the lua script?
moreover,can you develop the cache update interface by load cache dynamiclly, like bind9 rndc. thanks, a lot .
@vcunat
Petr Špaček
@pspacek
@beckhamaaa Hi. You have so many and so complex questions that it is better to discuss them on mailing list https://lists.nic.cz/mailman/listinfo/knot-resolver-users .
Some things, like your particular use-case for cache insert etc. might require too much time for free support - in that case you might be better off with paid support - please see https://www.knot-resolver.cz/support/
beckhamaaa
@beckhamaaa
oh, you should support my advice for free, because of a opensource. ok? thanks.
@pspacek
Tomas Krizek
@tomaskrizek
@micah_gitlab You're probably running into this issue with dig https://gitlab.isc.org/isc-projects/bind9/issues/1028 Either use newer version of dig, or unload the refuse_nord module that's turned on by default https://knot-resolver.readthedocs.io/en/stable/modules.html#refuse-queries-without-rd-bit
Tomas Krizek
@tomaskrizek
@beckhamaaa This is not what open source means. Our code is freely available under GNU GPLv3, which doesn't provide any guarantees or warranties. It also doesn't imply you're entitled to a free advice, that is only the developers' good will. If you actually want to know what is open source (or free software, in this case), read the actual license - https://gitlab.labs.nic.cz/knot/knot-resolver/blob/master/COPYING
beckhamaaa
@beckhamaaa
ok thanks i should know the definition of open source
@tomaskrizek
how can i get the level of you ? i a little know the udp network programming
Petr Špaček
@pspacek
What do you mean?
beckhamaaa
@beckhamaaa
how can enhance my develop kresd ability, i understand a little unit network progamming.
@pspacek
beckhamaaa
@beckhamaaa
docker run -v "$PWD:/data:rw" registry.labs.nic.cz/knot/shotgun/pellet:20191023 /data/captured.pcap -o /data/pellets.pcap -c 50000 -t 30 -r $RESOLVER_IP
where are the shotgun command parameter help ? i can't understand
@vcunat
Tomas Krizek
@tomaskrizek
@beckhamaaa There's --help and https://gitlab.labs.nic.cz/knot/shotgun/blob/master/README.md It's far from perfect and doesn't explain everything, but it's a start. I'd like to improve the docs and perhaps add some FAQ later.
beckhamaaa
@beckhamaaa
ok, i see ,thanks a lot , i have achieved 700 thousands qps in by tcpreplay.
@tomaskrizek
improve docs is a huge workload, but the idea is good ,support you
Robert Šefr
@robcza
@beckhamaaa what is the hardware used for 700k qps?
Robert Šefr
@robcza
@pspacek regarding the answer for cdn.analyzeo.com - I think that it is utter nonsense that the DNS client start recursive resolution on its own, though there is Authoritative section only in the response. The client asked with Recursion desired bit in the first place.
beckhamaaa
@beckhamaaa
10gbps nic
@robcza
Robert Šefr
@robcza
@beckhamaaa and cpu? cores and model?
beckhamaaa
@beckhamaaa
how can i insert the 1 billion RR to the cache quickly by lua script? such as the function : c:insert(rr, nil, ffi.C.KR_RANK_AUTH + ffi.C.KR_RANK_INSECURE)
@robcza
my cpu is 96 core,
Petr Špaček
@pspacek
@beckhamaaa I do not understand your question, just put that insert in a loop and go on.
Of course you need to have cache of sufficient size, and there might be some limit on maximum transaction size so it might be beneficial to split insert into smaller commits.
Limits would be imposed by LMDB library so you might need to read their docs.
beckhamaaa
@beckhamaaa
ok, thanks a lot, i have solved the insert speed question.
how can i load multi data.mdb by kresd daemon?
@pspacek
Vladimír Čunát
@vcunat
You can't. Not at once, at least. You can always re-open cache from a different directory, e.g. switch back and forth.
beckhamaaa
@beckhamaaa
thanks, i can't understand your meaning, that is to say, i can switch different data.mdb by kresd command line.
there are the nic link aggregation by you?
@vcunat
Vladimír Čunát
@vcunat
No, but I think such things are usually done at operating system level.
beckhamaaa
@beckhamaaa
ok ,can virtualization by exsi