Re: Multiple masters expected behavior?
\On 07/26/10 23:02, Barry Margolin wrote: In article, "Laws, Peter C." wrote: Understood, but what I'm asking about is that the slave does not appear to be losing contact with the first-listed master. In fact, from the logs, it appears to be flipping back and forth (though not round-robinning). Multiple masters is not about losing contact, it's about getting the most up-to-date version of the zone. There's no reason for the slave to A HA! So the answer to my original question, after all this, is "Yes" (this is expected behavior). Thanks. -- Peter Laws / N5UWY National Weather Center / Network Operations Center University of Oklahoma Information Technology pl...@ou.edu --- Feedback? Contact my director, Craig Cochell, cra...@ou.edu. Thank you! ___ bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
Re: Multiple masters expected behavior?
In article , "Laws, Peter C." wrote: > Understood, but what I'm asking about is that the slave does not appear to be > losing contact with the first-listed master. In fact, from the logs, it > appears to be flipping back and forth (though not round-robinning). Multiple masters is not about losing contact, it's about getting the most up-to-date version of the zone. There's no reason for the slave to assume that the first master has the best version of the zone. The only way to tell is to check the SOA records on all the masters, and perform a zone transfer from any of them that have a higher serial than the one you already have. -- Barry Margolin, bar...@alum.mit.edu Arlington, MA *** PLEASE don't copy me on replies, I'll read them in the group *** ___ bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
Re: Multiple masters expected behavior?
On 2010-07-23 22:52, Peter Laws wrote: > I would have expected that it would only ask the second-listed master if > the first didn't answer ... but I didn't write the code (and haven't > read it either! And how would your slave ever pick up an update on "second-listed master" that (for whatever reason) doesn't propagate to "the first"? After all, "the first" is still answering, but with old data. ___ bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
Re: Multiple masters expected behavior?
It makes it really hard to follow the thread. > Why not? > > Please don't top post! > From: "Laws, Peter C." > Date: Sun, 25 Jul 2010 16:56:26 + > Sender: bind-users-bounces+oberman=es@lists.isc.org > > Well aware of that, but we have RedHat support so we're stuck with > that given that the alternatives are self-supporting BIND (which you > could argue I'm doing right now!) or going with a 3rd party. Given > the economy, I'm pleased we're keeping RH support. While all of our (public) servers run on FreeBSD which has not shipped with 9.3 for a long time, we always run a near-current ISC release of BIND. The amount of support needed is trivial and I sleep much better at night that way. Yes, depending on the integration of your back-office DNS management/DNSSEC, it might be less so for some. Keeping the support of BIND on our public servers mostly unrelated to the IPAM and DNSSEC stuff has really not been hard. In the time it took me to send my reply, I could have updated BIND on all of our public servers and I don't have to upgrade all that often. I think running 9.3 is false economy. DNS is just too important. -- R. Kevin Oberman, Network Engineer Energy Sciences Network (ESnet) Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab) E-mail: ober...@es.net Phone: +1 510 486-8634 Key fingerprint:059B 2DDF 031C 9BA3 14A4 EADA 927D EBB3 987B 3751 ___ bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
RE: Multiple masters expected behavior?
Well aware of that, but we have RedHat support so we're stuck with that given that the alternatives are self-supporting BIND (which you could argue I'm doing right now!) or going with a 3rd party. Given the economy, I'm pleased we're keeping RH support. -- Peter Laws / N5UWY National Weather Center / Network Operations Center / Web University of Oklahoma Information Technology pl...@ou.edu From: Doug Barton [do...@dougbarton.us] Sent: Friday, July 23, 2010 19:23 To: Laws, Peter C. Cc: bind-us...@isc.org Subject: Re: Multiple masters expected behavior? On Thu, 22 Jul 2010, Peter Laws wrote: > BIND 9.3.6-P1-RedHat-9.3.6-4.P1.el5_4.2 9.3.x has been EOL for a long time now, FYI. -- Improve the effectiveness of your Internet presence with a domain name makeover!http://SupersetSolutions.com/ Computers are useless. They can only give you answers. -- Pablo Picasso ___ bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
RE: Multiple masters expected behavior?
Understood, but what I'm asking about is that the slave does not appear to be losing contact with the first-listed master. In fact, from the logs, it appears to be flipping back and forth (though not round-robinning). Someone else asked, essentially, "why?" ... The network paths are diverse to the different interfaces so, while I'm not protecting against failure of the master, I am protecting against network path failure. -- Peter Laws / N5UWY National Weather Center / Network Operations Center / Web University of Oklahoma Information Technology pl...@ou.edu From: bind-users-bounces+plaws=ou@lists.isc.org [bind-users-bounces+plaws=ou@lists.isc.org] on behalf of Barry Margolin [bar...@alum.mit.edu] Sent: Saturday, July 24, 2010 07:09 To: comp-protocols-dns-b...@isc.org Subject: Re: Multiple masters expected behavior? In article , Peter Laws wrote: > On 07/22/10 19:57, Barry Margolin wrote: > > In article, > > Peter Laws wrote: > > > >> I have multiple interfaces on my master and multiple interfaces on most of > >> my slaves. > >> > > > >> > >> Is that expected behavior? > > > > Yes. What if the first server stops getting updates, but the second one > > does and has a higher serial number? Don't you want the slaves to check > > the SOA record on it to pick up these changes? > > Except that the 2 "masters" are simply different interfaces on the same > master ... so the serial number *better* always be the same! That's true in *your* case. But BIND was designed to handle the more general case, where the masters can be different machines. -- Barry Margolin, bar...@alum.mit.edu Arlington, MA *** PLEASE don't copy me on replies, I'll read them in the group *** ___ bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users ___ bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
Re: Multiple masters expected behavior?
In article , Peter Laws wrote: > On 07/22/10 19:57, Barry Margolin wrote: > > In article, > > Peter Laws wrote: > > > >> I have multiple interfaces on my master and multiple interfaces on most of > >> my slaves. > >> > > > >> > >> Is that expected behavior? > > > > Yes. What if the first server stops getting updates, but the second one > > does and has a higher serial number? Don't you want the slaves to check > > the SOA record on it to pick up these changes? > > Except that the 2 "masters" are simply different interfaces on the same > master ... so the serial number *better* always be the same! That's true in *your* case. But BIND was designed to handle the more general case, where the masters can be different machines. -- Barry Margolin, bar...@alum.mit.edu Arlington, MA *** PLEASE don't copy me on replies, I'll read them in the group *** ___ bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
Re: Multiple masters expected behavior?
On Thu, 22 Jul 2010, Peter Laws wrote: BIND 9.3.6-P1-RedHat-9.3.6-4.P1.el5_4.2 9.3.x has been EOL for a long time now, FYI. -- Improve the effectiveness of your Internet presence with a domain name makeover!http://SupersetSolutions.com/ Computers are useless. They can only give you answers. -- Pablo Picasso ___ bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
Re: Multiple masters expected behavior?
On Fri, 23 Jul 2010, Peter Laws wrote: Except that the 2 "masters" are simply different interfaces on the same master Why do you think that would be helpful? Or are you just testing the multi-master configuration in the hopes of adding actual diversity down the road? Doug -- Improve the effectiveness of your Internet presence with a domain name makeover!http://SupersetSolutions.com/ Computers are useless. They can only give you answers. -- Pablo Picasso ___ bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
Re: Multiple masters expected behavior?
On 07/22/10 19:57, Barry Margolin wrote: In article, Peter Laws wrote: I have multiple interfaces on my master and multiple interfaces on most of my slaves. Is that expected behavior? Yes. What if the first server stops getting updates, but the second one does and has a higher serial number? Don't you want the slaves to check the SOA record on it to pick up these changes? Except that the 2 "masters" are simply different interfaces on the same master ... so the serial number *better* always be the same! Looking at the logs, it appears that the choice of masters is a second-to-second thing because what I'm seeing is that one zone goes via one interface and then the next zone, perhaps only a few 10s of ms later, goes via the other interface. I would have expected that it would only ask the second-listed master if the first didn't answer ... but I didn't write the code (and haven't read it either! -- Peter Laws / N5UWY National Weather Center / Network Operations Center University of Oklahoma Information Technology pl...@ou.edu --- Feedback? Contact my director, Craig Cochell, cra...@ou.edu. Thank you! ___ bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
Re: Multiple masters expected behavior?
In article , Peter Laws wrote: > I have multiple interfaces on my master and multiple interfaces on most of > my slaves. > > I've got one of the slaves set up so that its masters {}; statement has two > of the master's interfaces in it. The preferred is first, with the > non-preferred second. I was contemplating using this on all slaves to > guard against a network path failure. > > Note that I also have both of the slave's interfaces in the also-notify > statement on the master (it's an unpublished slave). > > I would have thought that BIND would always hit the first and never the > second. That doesn't seem to be the case however. In fact, in a few cases > I've seen it seems to use both, though not round-robinning that I can see > from the logs. > > Is that expected behavior? Yes. What if the first server stops getting updates, but the second one does and has a higher serial number? Don't you want the slaves to check the SOA record on it to pick up these changes? -- Barry Margolin, bar...@alum.mit.edu Arlington, MA *** PLEASE don't copy me on replies, I'll read them in the group *** ___ bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
Re: Multiple masters expected behavior?
On 07/22/2010 10:59 PM, Peter Laws wrote: I have multiple interfaces on my master and multiple interfaces on most of my slaves. I've got one of the slaves set up so that its masters {}; statement has two of the master's interfaces in it. The preferred is first, with the non-preferred second. I was contemplating using this on all slaves to guard against a network path failure. Note that I also have both of the slave's interfaces in the also-notify statement on the master (it's an unpublished slave). I would have thought that BIND would always hit the first and never the second. That doesn't seem to be the case however. In fact, in a few cases I've seen it seems to use both, though not round-robinning that I can see from the logs. I believe like all DNS servers, bind will pick the quickest-responding one (with the highest SOA serial, of course). It will certainly send SOA queries to both in case one master has a higher serial than the other. ___ bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
Multiple masters expected behavior?
I have multiple interfaces on my master and multiple interfaces on most of my slaves. I've got one of the slaves set up so that its masters {}; statement has two of the master's interfaces in it. The preferred is first, with the non-preferred second. I was contemplating using this on all slaves to guard against a network path failure. Note that I also have both of the slave's interfaces in the also-notify statement on the master (it's an unpublished slave). I would have thought that BIND would always hit the first and never the second. That doesn't seem to be the case however. In fact, in a few cases I've seen it seems to use both, though not round-robinning that I can see from the logs. Is that expected behavior? BIND 9.3.6-P1-RedHat-9.3.6-4.P1.el5_4.2 -- Peter Laws / N5UWY National Weather Center / Network Operations Center University of Oklahoma Information Technology pl...@ou.edu --- Feedback? Contact my director, Craig Cochell, cra...@ou.edu. Thank you! ___ bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users