At 02:31 PM 2/26/97 -0500, Bradley Dunn wrote:
On Wed, 26 Feb 1997, Lyndon Levesley wrote:
Nameservers are a bit harder to renumber, but that's not too bad.
When you have hundreds of virtual web sites?
Uhm, sure. Just slowly decrease the TTL on your name server until you are ready to renumber (and the TTL is set to something ridiculously low such as 1 minute). The customers will then experience something like a 1 minute outage when you renumber. If we are going to start getting into the procedures of HOW to renumber this should likely move to the PIER mailing list. I think if in general people become more interested in HOW to gracefully renumber themselves and their customers instead of worrying about how hard it is to do they would see that while it is work, it isn't really all that hard. Justin Newton Network Architect Erol's Internet Services ISP/C Director at Large
At 02:31 PM 2/26/97 -0500, Bradley Dunn wrote:
On Wed, 26 Feb 1997, Lyndon Levesley wrote:
Nameservers are a bit harder to renumber, but that's not too bad.
When you have hundreds of virtual web sites?
Uhm, sure. Just slowly decrease the TTL on your name server until you are ready to renumber (and the TTL is set to something ridiculously low such as 1 minute). The customers will then experience something like a 1 minute outage when you renumber. If we are going to start getting into the procedures of HOW to renumber this should likely move to the PIER mailing list. I think if in general people become more interested in HOW to gracefully renumber themselves and their customers instead of worrying about how hard it is to do they would see that while it is work, it isn't really all that hard.
Justin Newton Network Architect Erol's Internet Services ISP/C Director at Large
You're making lots of assumptions. 1) That client DNS systems will actually honor such a TTL. Many don't (claim they're broken all you want, but these are the facts). 2) That client SOFTWARE will actually go back and ask again for the IP number. Several won't (Netscrape being rumored to be one of them). TTLs are irrelavent in that case. Go ahead and try to tell your customer, who purchased web service from you, that you have the right to disrupt their operations at any time and under any pretense and see how many of them you have left. -- -- Karl Denninger (karl@MCS.Net)| MCSNet - The Finest Internet Connectivity http://www.mcs.net/~karl | T1's from $600 monthly to FULL DS-3 Service | 99 Analog numbers, 77 ISDN, Web servers $75/mo Voice: [+1 312 803-MCS1 x219]| Email to "info@mcs.net" WWW: http://www.mcs.net/ Fax: [+1 312 803-4929] | 2 FULL DS-3 Internet links; 400Mbps B/W Internal
Karl Denninger supposedly said:
You're making lots of assumptions.
1) That client DNS systems will actually honor such a TTL. Many don't (claim they're broken all you want, but these are the facts).
2) That client SOFTWARE will actually go back and ask again for the IP number. Several won't (Netscrape being rumored to be one of them). TTLs are irrelavent in that case.
Go ahead and try to tell your customer, who purchased web service from you, that you have the right to disrupt their operations at any time and under any pretense and see how many of them you have left.
Karl, How do you handle hardware upgrades, random crashes, etc. with your clients? Do you give them a refund for such downtimes? DO you guarentee that every client that tries to access their web page will always get through? My guess is you don't. You perform a service for them and probably schedule maintenence in such a way as to minimize downtime and impact on that service. If you have a better scheme, like fully redundent machines that fall over automatically and let you do maintenence on one while the other opperates then I think you have done an excellent job at providing a quality service for your customers. On the other hand, someone who has done such a setup should realize how easy it would be to migrate it to different addresses while maintaining pretty much complete connectivity for the old addresses for a reasonable time (like a standard TTL length). ---> Phil
Karl Denninger supposedly said:
You're making lots of assumptions.
1) That client DNS systems will actually honor such a TTL. Many don't (claim they're broken all you want, but these are the facts).
2) That client SOFTWARE will actually go back and ask again for the IP number. Several won't (Netscrape being rumored to be one of them). TTLs are irrelavent in that case.
Go ahead and try to tell your customer, who purchased web service from you, that you have the right to disrupt their operations at any time and under any pretense and see how many of them you have left.
Karl,
How do you handle hardware upgrades, random crashes, etc. with your clients? Do you give them a refund for such downtimes? DO you guarentee that every client that tries to access their web page will always get through?
My guess is you don't. You perform a service for them and probably schedule maintenence in such a way as to minimize downtime and impact on that service.
If you have a better scheme, like fully redundent machines that fall over automatically and let you do maintenence on one while the other opperates then I think you have done an excellent job at providing a quality service for your customers. On the other hand, someone who has done such a setup should realize how easy it would be to migrate it to different addresses while maintaining pretty much complete connectivity for the old addresses for a reasonable time (like a standard TTL length).
---> Phil
We have a much better scheme. Try coordinating multiple servers on multiple addresses, against the same document and log data, without any possibility of corruption during a significant period of time. Second, why should we have to have twice the infrastructure in place for such an event? But let's put that aside for a minute, because it isn't the biggest issue. Its not our internal infrastructure that is the worst problem (although that's certainly significant and a royal pain in the ass). It can be argued, though, that the disruption involved there comes about as a consequence of our decision, and it impacts us. That's "fair", especially if everyone has to play by the same rules (ie: an MCI customer who leaves has to renumber, an MCSNet customer who leaves has to renumber, etc). But at the indirect level its a different issue entirely. Our CUSTOMER'S networks should not be subject to disruption because *WE*, as their IP packet supplier, change a vendor relationship. Again, if you don't know why this is the case, then you need to talk to some attorneys. Tying customers like this at an indirect level can expose you to all kinds of fun and games, none of them technical. Building policy *as a group* which acts to restrain trade (and this most certainly does) is extraordinarily dangerous. I will have nothing to do with any association or group which does this, because from my analysis the risk exists that such an act is a criminal violation of US Federal law. -- -- Karl Denninger (karl@MCS.Net)| MCSNet - The Finest Internet Connectivity http://www.mcs.net/~karl | T1's from $600 monthly to FULL DS-3 Service | 99 Analog numbers, 77 ISDN, Web servers $75/mo Voice: [+1 312 803-MCS1 x219]| Email to "info@mcs.net" WWW: http://www.mcs.net/ Fax: [+1 312 803-4929] | 2 FULL DS-3 Internet links; 400Mbps B/W Internal
Justin,
You're making lots of assumptions.
1) That client DNS systems will actually honor such a TTL. Many don't (claim they're broken all you want, but these are the facts).
2) That client SOFTWARE will actually go back and ask again for the IP number. Several won't (Netscrape being rumored to be one of them). TTLs are irrelavent in that case.
Go ahead and try to tell your customer, who purchased web service from you, that you have the right to disrupt their operations at any time and under any pretense and see how many of them you have left.
Leave the site running on both IPs for (say) 4 weeks. Even broken DNS doesn't last this long. And yes, anyone who doesn't quit their web browser for four weeks will suffer, but these are mostly UNIX guys anyway who will know why it's broken :-) Alex Bligh Xara Networks
"Justin W. Newton" wrote : |-> At 02:31 PM 2/26/97 -0500, Bradley Dunn wrote: |-> >On Wed, 26 Feb 1997, Lyndon Levesley wrote: |-> > |-> >> Nameservers are a bit harder to renumber, but that's not too bad. |-> > |-> >When you have hundreds of virtual web sites? |-> |-> Uhm, sure. Just slowly decrease the TTL on your name server until you are |-> ready to renumber (and the TTL is set to something ridiculously low such as |-> 1 minute). The customers will then experience something like a 1 minute |-> outage when you renumber. If we are going to start getting into the |-> procedures of HOW to renumber this should likely move to the PIER mailing |-> list. I think if in general people become more interested in HOW to |-> gracefully renumber themselves and their customers instead of worrying |-> about how hard it is to do they would see that while it is work, it isn't |-> really all that hard. |-> Without wanting to get into pier territory too much, this relies on applications and networks not having DNS caches. Caches do get in the way, particularly if an application is doing the caching ;( ) That said, renumbering is made a lot easier if you know from day 1 that it is inevitable. (My way of renumbering nameservers, BTW, involves a "changeover time" where basically you borrow the old address space for a while and give your nameserver both the old and new address. You then profile you traffic until noone hits the old address, at which point you remove it.) |-> Justin Newton |-> Network Architect |-> Erol's Internet Services |-> ISP/C Director at Large |-> Lyndon Levesley Xara Networks I've had a wonderful time... ...but this wasn't it.
TTL is meaningless because Navigator (and probably MSIE and every other browser) caches DNS "forever". There are people who don't restart their browsers every day... as I mentioned in another message, I've seen people going to the old ips up to three weeks after the theoretical time after which they should have changed. Dean On Wed, 26 Feb 1997, Justin W. Newton wrote:
At 02:31 PM 2/26/97 -0500, Bradley Dunn wrote:
On Wed, 26 Feb 1997, Lyndon Levesley wrote:
Nameservers are a bit harder to renumber, but that's not too bad.
When you have hundreds of virtual web sites?
Uhm, sure. Just slowly decrease the TTL on your name server until you are ready to renumber (and the TTL is set to something ridiculously low such as 1 minute). The customers will then experience something like a 1 minute outage when you renumber. If we are going to start getting into the procedures of HOW to renumber this should likely move to the PIER mailing list. I think if in general people become more interested in HOW to gracefully renumber themselves and their customers instead of worrying about how hard it is to do they would see that while it is work, it isn't really all that hard.
Justin Newton Network Architect Erol's Internet Services ISP/C Director at Large
On Wed, 26 Feb 1997, Dean Gaudet wrote:
TTL is meaningless because Navigator (and probably MSIE and every other browser) caches DNS "forever". There are people who don't restart their [...]
Do you (or anybody else) know the behavior of the popular proxy caches out there WRT DNS TTLs and mutiple A records? -- Matt Ranney - mjr@ranney.com This is how I sign all my messages.
-----BEGIN PGP SIGNED MESSAGE----- There are ways to use both numbers simulatenously. Case in point, we have 3 sun boxes with IP addresses on their ethenet ports of 206.127.64.130,131 & 132. They used to be 204.94.230.2, 3, & 4. Through the wonders of vif's and host routes BOTH IP numbers are still reachable and usable for each box. We had to do this because some of our dialup customers (wich we are RENUMBERING to "dynamic") are still pointing to 2 & 3 as thier DNS addresses. BTW we havent't lost a customer. The real trick is to only do it once. DON't renumber until you can justify an /18 or so from the internic. I'm sure the internic will be more than willing to provide you with an /18 to renumber into if you have close to an /18 of space in use. Just buckle down and say with your upstream until then. - -forrestc@imach.com On Wed, 26 Feb 1997, Justin W. Newton wrote:
At 02:31 PM 2/26/97 -0500, Bradley Dunn wrote:
On Wed, 26 Feb 1997, Lyndon Levesley wrote:
Nameservers are a bit harder to renumber, but that's not too bad.
When you have hundreds of virtual web sites?
Uhm, sure. Just slowly decrease the TTL on your name server until you are ready to renumber (and the TTL is set to something ridiculously low such as 1 minute). The customers will then experience something like a 1 minute outage when you renumber. If we are going to start getting into the procedures of HOW to renumber this should likely move to the PIER mailing list. I think if in general people become more interested in HOW to gracefully renumber themselves and their customers instead of worrying about how hard it is to do they would see that while it is work, it isn't really all that hard.
Justin Newton Network Architect Erol's Internet Services ISP/C Director at Large
-----BEGIN PGP SIGNATURE----- Version: 2.6.2 iQCVAwUBMxTakez4AI4yZZixAQErGwP+JcK4HwNKkHUhntt1vLLRv7S+EbT1W7qB RFKPteVg8aYX3STmfu/wFldHGtnWlJAycVnDd9iDZjsavEbzpGeXOCjJjqOon5Kv s/8Uujm8VJdasx8wxgjBvdE+ZzMnIzYYBn2d/67pRbpRHhASF2ZsbN1R8owAcmgv 77wjQQnBxsI= =hZrH -----END PGP SIGNATURE-----
participants (8)
-
Alex.Bligh
-
Dean Gaudet
-
Forrest W. Christian
-
Justin W. Newton
-
Karl Denninger
-
Lyndon Levesley
-
Matt Ranney
-
Philip J. Nesser II