At 10:47 PM 1/6/2003 +0000, you wrote: >On Mon, Jan 06, 2003 at 04:42:42PM +0000, Beast wrote: >> 1. caching only ns jangan diinstall dlm 1 ip (1 mesin boleh, lain >> ip), kenapa? > >maksudnya barangkali: caching dns (dnscache) tidak bisa listen di ip >yang sama dengan authoritative dns (tinydns). karena, walaupun >tinydns hanya menggunakan UDP, dnscache menggunakan baik TCP maupun >UDP, jadi kalau satu sudah jalan, yang lain akan mendapat error >EADDRINUSE (Address already in use or something like that).
http://cr.yp.to/djbdns/separation.html > >> 2. di clint's /etc/resolve.conf, jangan point langsung ke primary ns, tapi >> ke caching only ns, betulkah? > >betul, kalau maksud dari 'primary ns' adalah authoritative dns >(tinydns), karena tinydns tidak melakukan recursive query, jadi >adalah hil yang mustahal kalau bisa resolve other domain dari >tinydns. > >> dgn bind9, apakah "rule" diatas perlu? bagaimana jika install saja >> primary svr dan sec , dan user point langsung ke server2 tsb? > >pertama, dalam dunia dns, tidak ada istilah primary/secondary itu >murni istilah administratif. kedua, biasanya istilah itu dikaitkan >dengan istilah 'master/slave' di bind. yang ada adalah caching dns >(resolver), authoritative dns. Iya, sorry :-) seharusnya Authoritative servers dan caching server. > >tapi kalau penangkapan saya terhadap pertanyaan di atas benar, >jawabnya tidak perlu. bind, by default, menjalankan >caching/authoritative nameserver di 'satu' daemon. tanda petik >penting, karena bind9 pakai thread, jadi tidak bisa dibilang satu >daemon lagi :-) walaupun ini configurable saat compile. Ini yg saya cari, tapi bener kalau bertindakn sbg authoritative nameserver, bertindak juga sbg caching server? Tapi dalam case saya (internal dns) yg hanya melayani 1 domain sepertinya ndak perlu, hanya tinggal set master dan lainya slave. Tks. ini mugkin berguna buat yg lain : DNS performs two types of queries (or lookups) iterative and recursive. The description assumes no cache exists. Iterative queries are "normally" used between DNS servers. An iterative query request tells the DNS server that receives the request that it only needs to provide as much information as it has and that is it. So when you query the root servers they only give you a pointer to the DNS server that supports the level 1 domain you are interested (such as com. edu. org. etc...). The requesting DNS server will then sends its query to the level 1 domain DNS server, it will again give you a pointer down to the level 2 DNS server, and so on and so on. The DNS server needing the information is finding the information by sending several iterations of the same query. Recursive queries are different. Essentially what happens is the sender of the query is saying to a DNS server, "find me the answer to this question, or send me an error back". But the DNS server that accepts a recursive query request assumes the responsibility to find the answer to the question presented in the query or return the appropriate error. So the sender of the recursive query only sends a single query to a DNS server, and will get its answer back. Recursive queries are generally sent from stub resolvers and DNS servers configured to forward. -- Right or wrong my list. Unsubscribe option is currently unavailable. Indeed, it's available upon request .. but: cepek dulu donk!
