[Pdns-users] pdns recursor - unthreaded vs threaded operation
Hello list, I'm a newbie of pdns and I'm setting up a recursor-only installation as local dns cache for a high load smtp server. I'm using the lua scripting feature as a way to track down MX queries. In particular, the preresolve() function relies on luasql to execute queries on a (local) mysql database. This is my problem: if I enable more than one thread in recursor.conf I get the lua script loaded more than once, and ultimately concurrent queries to the db that could mess up the stored data, depending on the execution order of select and update queries by the various threads. I'd like to avoid circumventing these race conditions through complicated lua coding, and instead I'd prefer to keep things as simple as possible and limit the number of pdns threads to just 1 (threads=1 setting in recursor.conf). I know that the lua feature is not meant to be used this way, but this could really help me solving a big problem :) Now my questions are: am I right thinking this would solve my problem? how deep is the impact on performances changing the number of threads from 2 (default setting) to just 1? are there other sensible settings that I could adjust to better suit my needs? Thanks a lot for your support. Cheers, Fabio ___ Pdns-users mailing list Pdns-users@mailman.powerdns.com http://mailman.powerdns.com/mailman/listinfo/pdns-users
[Pdns-users] powerdns 3.0 notify not logged
Hi All, I dont know if this should be a bug or a feature: We use powerdns for a long time as slaves, the master are bind servers that notify updates to the powerdns servers. When the bind notify the updated serial, that notify should be loggen in the powerdns log. But since 3.0, whatever i try i cannot get that logging line back. I tried almost every possible option for logging, but that one looks disappeared from logging. Anyone that confirm this one, or knows a solution ? Thanks, Mike ___ Pdns-users mailing list Pdns-users@mailman.powerdns.com http://mailman.powerdns.com/mailman/listinfo/pdns-users
Re: [Pdns-users] Bad cache TTL with pdns-resursor
Hi Michael, The SOA will indeed probably not go to zero because it will have been renewed in that time. PowerDNS Recursor 3.3 and later take care to decrement the TTL when serving contents from the packet cache. So when you don't see it going down in 3.2, that is correct, and maybe a bit strange, but not a bug. Bert On Thu, Dec 29, 2011 at 10:08:53AM +0100, Michael FROMENT wrote: Hi, I have some strange cached TTL on my pdns-recursor ( version 3.2-4 under debian ). When I query my recursor the TTL doesn't decrease by 1 second. Moreover the TTL never goes to 0 so the zone never expire in cache. As an example: root@BBK2NS:~# dig soa google.com @89.2.0.1 ;; ANSWER SECTION: google.com. 569 IN SOA ns1.google.com. dns-admin.google.com. 1472387 7200 1800 1209600 300 ;; Query time: 0 msec ;; SERVER: 89.2.0.1#53(89.2.0.1) ;; WHEN: Thu Dec 29 10:02:46 2011 ;; MSG SIZE rcvd: 78 root@BBK2NS:~# dig soa google.com @89.2.0.1 ;; ANSWER SECTION: google.com. 40 IN SOA ns1.google.com. dns-admin.google.com. 1472388 7200 1800 1209600 300 ;; Query time: 0 msec ;; SERVER: 89.2.0.1#53(89.2.0.1) ;; WHEN: Thu Dec 29 10:02:47 2011 ;; MSG SIZE rcvd: 78 root@BBK2NS:~# dig soa google.com @89.2.0.1 ;; ANSWER SECTION: google.com. 562 IN SOA ns1.google.com. dns-admin.google.com. 1472388 7200 1800 1209600 300 ;; Query time: 0 msec ;; SERVER: 89.2.0.1#53(89.2.0.1) ;; WHEN: Thu Dec 29 10:04:08 2011 ;; MSG SIZE rcvd: 78 Is this a known bug on the recursor? Kind Regards Michael ___ Pdns-users mailing list Pdns-users@mailman.powerdns.com http://mailman.powerdns.com/mailman/listinfo/pdns-users ___ Pdns-users mailing list Pdns-users@mailman.powerdns.com http://mailman.powerdns.com/mailman/listinfo/pdns-users
Re: [Pdns-users] Bad cache TTL with pdns-resursor
Hi bert, Thanks for the answer I've uderstand the strange ttl response when the answer come from the paquet cache. Now here is a little more strange problem ;; QUESTION SECTION: ;s4.noelshack.com. IN A ;; ANSWER SECTION: s4.noelshack.com. 3600IN CNAME noelshack.com. noelshack.com. 3600IN A 91.121.63.175 ;; QUESTION SECTION: ;s4.noelshack.com. IN NS ;; ANSWER SECTION: s4.noelshack.com. 2175IN CNAME noelshack.com. noelshack.com. 85974 IN NS ns1.afraid.org. noelshack.com. 85974 IN NS ns4.afraid.org. noelshack.com. 85974 IN NS ns2.afraid.org. noelshack.com. 85974 IN NS ns3.afraid.org. ;; QUESTION SECTION: ;noelshack.com. IN SOA ;; ANSWER SECTION: noelshack.com. 3535IN SOA ns1.afraid.org. dnsadmin.afraid.org. 1110170003 86400 7200 2419200 3600 The zone noelshack.com. have moved since more than a week at ns-01.odysseeinteractive.com The correct answer from another pdns_recursor : # ;; QUESTION SECTION: ;noelshack.com. IN SOA ;; ANSWER SECTION: noelshack.com. 86281 IN SOA ns-01.odysseeinteractive.com. noc.odysseeinteractive.com. 2011122801 86400 7200 604800 86400 # The ttl on this zone nerver goes to 0 and moreover after 1 hour ( the paquet cache duration) is sometime increased. We have around 100 servers and half are affected by this problem... The only solution I've find is to wipe the cache for this dns entrie. There is no reason to do this since 1 week as the TTL on the soa is only 86400... Any ideas? Kind regards Michael On 12/29/2011 04:20 PM, bert hubert wrote: Hi Michael, The SOA will indeed probably not go to zero because it will have been renewed in that time. PowerDNS Recursor 3.3 and later take care to decrement the TTL when serving contents from the packet cache. So when you don't see it going down in 3.2, that is correct, and maybe a bit strange, but not a bug. Bert On Thu, Dec 29, 2011 at 10:08:53AM +0100, Michael FROMENT wrote: Hi, I have some strange cached TTL on my pdns-recursor ( version 3.2-4 under debian ). When I query my recursor the TTL doesn't decrease by 1 second. Moreover the TTL never goes to 0 so the zone never expire in cache. As an example: root@BBK2NS:~# dig soa google.com @89.2.0.1 ;; ANSWER SECTION: google.com. 569 IN SOA ns1.google.com. dns-admin.google.com. 1472387 7200 1800 1209600 300 ;; Query time: 0 msec ;; SERVER: 89.2.0.1#53(89.2.0.1) ;; WHEN: Thu Dec 29 10:02:46 2011 ;; MSG SIZE rcvd: 78 root@BBK2NS:~# dig soa google.com @89.2.0.1 ;; ANSWER SECTION: google.com. 40 IN SOA ns1.google.com. dns-admin.google.com. 1472388 7200 1800 1209600 300 ;; Query time: 0 msec ;; SERVER: 89.2.0.1#53(89.2.0.1) ;; WHEN: Thu Dec 29 10:02:47 2011 ;; MSG SIZE rcvd: 78 root@BBK2NS:~# dig soa google.com @89.2.0.1 ;; ANSWER SECTION: google.com. 562 IN SOA ns1.google.com. dns-admin.google.com. 1472388 7200 1800 1209600 300 ;; Query time: 0 msec ;; SERVER: 89.2.0.1#53(89.2.0.1) ;; WHEN: Thu Dec 29 10:04:08 2011 ;; MSG SIZE rcvd: 78 Is this a known bug on the recursor? Kind Regards Michael ___ Pdns-users mailing list Pdns-users@mailman.powerdns.com http://mailman.powerdns.com/mailman/listinfo/pdns-users ___ Pdns-users mailing list Pdns-users@mailman.powerdns.com http://mailman.powerdns.com/mailman/listinfo/pdns-users
Re: [Pdns-users] Bad cache TTL with pdns-resursor
On Thu, Dec 29, 2011, Michael FROMENT wrote: Hi bert, Thanks for the answer I've uderstand the strange ttl response when the answer come from the paquet cache. Now here is a little more strange problem The zone noelshack.com. have moved since more than a week at ns-01.odysseeinteractive.com Hello Michael, (I was just reading the frnog related topic ;) ) this is a problem we already encounter here, due to a bug in version 3.2 of the recursor : http://wiki.powerdns.com/trac/changeset/1548 The cache still continues using old NS informations concerning the domain (as the old authoritative servers still answer for it), and so the migration of the domain to new authoritatives servers is not seen by PowerDNS Recursor. You should consider upgrading your version of the recursor, and if so, going directly to version 3.4-pre, because we also encounter another irritative bug in version 3.3: when an authoritative server wrongly announces itself as authoritative for another domain :) (especially domains hosted on ns20.ovh.net, that answers itself as authoritative for google.com ;) ). Kind regards, -- David Gavarret ___ Pdns-users mailing list Pdns-users@mailman.powerdns.com http://mailman.powerdns.com/mailman/listinfo/pdns-users
Re: [Pdns-users] Bad cache TTL with pdns-resursor
Hi David, Thanks for the bug related link and the solution... We will upgrade our recursor asap. Nice to meet you on another mailing list... as you can see we were on the problem Regards Michael On 12/29/2011 05:08 PM, GAVARRET, David wrote: On Thu, Dec 29, 2011, Michael FROMENT wrote: Hi bert, Thanks for the answer I've uderstand the strange ttl response when the answer come from the paquet cache. Now here is a little more strange problem The zone noelshack.com. have moved since more than a week at ns-01.odysseeinteractive.com Hello Michael, (I was just reading the frnog related topic ;) ) this is a problem we already encounter here, due to a bug in version 3.2 of the recursor : http://wiki.powerdns.com/trac/changeset/1548 The cache still continues using old NS informations concerning the domain (as the old authoritative servers still answer for it), and so the migration of the domain to new authoritatives servers is not seen by PowerDNS Recursor. You should consider upgrading your version of the recursor, and if so, going directly to version 3.4-pre, because we also encounter another irritative bug in version 3.3: when an authoritative server wrongly announces itself as authoritative for another domain :) (especially domains hosted on ns20.ovh.net, that answers itself as authoritative for google.com ;) ). Kind regards, ___ Pdns-users mailing list Pdns-users@mailman.powerdns.com http://mailman.powerdns.com/mailman/listinfo/pdns-users
Re: [Pdns-users] pdns recursor - unthreaded vs threaded operation
Fabio Sangiovanni wrote: I'm a newbie of pdns and I'm setting up a recursor-only installation as local dns cache for a high load smtp server. I'm using the lua scripting feature as a way to track down MX queries. In particular, the preresolve() function relies on luasql to execute queries on a (local) mysql database. This is my problem: if I enable more than one thread in recursor.conf I get the lua script loaded more than once, and ultimately concurrent queries to the db that could mess up the stored data, depending on the execution order of select and update queries by the various threads. I'd like to avoid circumventing these race conditions through complicated lua coding, and instead I'd prefer to keep things as simple as possible and limit the number of pdns threads to just 1 (threads=1 setting in recursor.conf). I know that the lua feature is not meant to be used this way, but this could really help me solving a big problem :) Now my questions are: am I right thinking this would solve my problem? how deep is the impact on performances changing the number of threads from 2 (default setting) to just 1? are there other sensible settings that I could adjust to better suit my needs? Wouldn't using database transactions solve your integrity problem? Nils. ___ Pdns-users mailing list Pdns-users@mailman.powerdns.com http://mailman.powerdns.com/mailman/listinfo/pdns-users