cgnat - how do you handle customer issues
Couple questions please. When you put thousands of customers behind a cgnat boundary, how do you all handle customer complaints about the following. 1 - for external connectivity to the customers premise devices, not being able to access web servers, web cameras, etc, in their premises? 2 - from the premise natted device, when customers go to a university or bank web site, how do you handle randomly changing ip addresses/ports that may occur due to idle time and session tear-down in nat table such that the bank website has issues with seeing your session ip change? -Aaron
I'm a fan of nailing each customer IP to a particular range of ports on a given public IP. Real easy to track who did what and to prevent shifting IPs. ----- Mike Hammett Intelligent Computing Solutions Midwest Internet Exchange The Brothers WISP ----- Original Message ----- From: "Aaron Gould" <aaron1@gvtc.com> To: Nanog@nanog.org Sent: Tuesday, February 27, 2018 10:30:21 AM Subject: cgnat - how do you handle customer issues Couple questions please. When you put thousands of customers behind a cgnat boundary, how do you all handle customer complaints about the following. 1 - for external connectivity to the customers premise devices, not being able to access web servers, web cameras, etc, in their premises? 2 - from the premise natted device, when customers go to a university or bank web site, how do you handle randomly changing ip addresses/ports that may occur due to idle time and session tear-down in nat table such that the bank website has issues with seeing your session ip change? -Aaron
For number 2, I'm a fan of what mike suggests. I believe the technical term is MAP-T. For number 1, anyone who wants one, gets one. We provide free public static IP to any customer who asks for one. Another solution, using above solution is to ask them which ports they need, and forward those to them using a port within their assign range. i.e. teach them how to access their home web server using a different port(say 32424, or similar). This won't solve all the issues, which is why we use solution 1. On 27 February 2018 at 09:32, Mike Hammett <nanog@ics-il.net> wrote:
I'm a fan of nailing each customer IP to a particular range of ports on a given public IP. Real easy to track who did what and to prevent shifting IPs.
----- Mike Hammett Intelligent Computing Solutions
Midwest Internet Exchange
The Brothers WISP
----- Original Message -----
From: "Aaron Gould" <aaron1@gvtc.com> To: Nanog@nanog.org Sent: Tuesday, February 27, 2018 10:30:21 AM Subject: cgnat - how do you handle customer issues
Couple questions please. When you put thousands of customers behind a cgnat boundary, how do you all handle customer complaints about the following.
1 - for external connectivity to the customers premise devices, not being able to access web servers, web cameras, etc, in their premises?
2 - from the premise natted device, when customers go to a university or bank web site, how do you handle randomly changing ip addresses/ports that may occur due to idle time and session tear-down in nat table such that the bank website has issues with seeing your session ip change?
-Aaron
Thanks For #2 – what if the ports allocated aren’t enough for the amount of inet traffic the customer site uses ? …is the customer denied service based on insufficient port range ? …or are they assigned another block within that some ip’s range of I think it’s 0-64k or 1025-64k… but how far can you take that before there aren’t anymore port blocks left on that single ip ? …and if you have to allocate that customer another port block from a *different* ip, then we are in the situation of the bank website not liking the fact that the session is bouncing to a different ip maybe ? - Aaron From: Michael Crapse [mailto:michael@wi-fiber.io] Sent: Tuesday, February 27, 2018 11:19 AM To: Mike Hammett Cc: Aaron Gould; NANOG list Subject: Re: cgnat - how do you handle customer issues For number 2, I'm a fan of what mike suggests. I believe the technical term is MAP-T. For number 1, anyone who wants one, gets one. We provide free public static IP to any customer who asks for one. Another solution, using above solution is to ask them which ports they need, and forward those to them using a port within their assign range. i.e. teach them how to access their home web server using a different port(say 32424, or similar). This won't solve all the issues, which is why we use solution 1. On 27 February 2018 at 09:32, Mike Hammett <nanog@ics-il.net> wrote: I'm a fan of nailing each customer IP to a particular range of ports on a given public IP. Real easy to track who did what and to prevent shifting IPs. ----- Mike Hammett Intelligent Computing Solutions Midwest Internet Exchange The Brothers WISP ----- Original Message ----- From: "Aaron Gould" <aaron1@gvtc.com> To: Nanog@nanog.org Sent: Tuesday, February 27, 2018 10:30:21 AM Subject: cgnat - how do you handle customer issues Couple questions please. When you put thousands of customers behind a cgnat boundary, how do you all handle customer complaints about the following. 1 - for external connectivity to the customers premise devices, not being able to access web servers, web cameras, etc, in their premises? 2 - from the premise natted device, when customers go to a university or bank web site, how do you handle randomly changing ip addresses/ports that may occur due to idle time and session tear-down in nat table such that the bank website has issues with seeing your session ip change? -Aaron
I utilize A10 CGNAT that allows dynamic NAT logging, since we're in a similar boat of utilization. This email has been sent from my phone. Please excuse any brevity, typos, or lack of formality. ________________________________ From: Aaron Gould <aaron1@gvtc.com> Sent: Tuesday, February 27, 2018 12:54 To: 'Michael Crapse'; 'Mike Hammett' Cc: 'NANOG list' Subject: RE: cgnat - how do you handle customer issues Thanks For #2 – what if the ports allocated aren’t enough for the amount of inet traffic the customer site uses ? …is the customer denied service based on insufficient port range ? …or are they assigned another block within that some ip’s range of I think it’s 0-64k or 1025-64k… but how far can you take that before there aren’t anymore port blocks left on that single ip ? …and if you have to allocate that customer another port block from a *different* ip, then we are in the situation of the bank website not liking the fact that the session is bouncing to a different ip maybe ? - Aaron From: Michael Crapse [mailto:michael@wi-fiber.io] Sent: Tuesday, February 27, 2018 11:19 AM To: Mike Hammett Cc: Aaron Gould; NANOG list Subject: Re: cgnat - how do you handle customer issues For number 2, I'm a fan of what mike suggests. I believe the technical term is MAP-T. For number 1, anyone who wants one, gets one. We provide free public static IP to any customer who asks for one. Another solution, using above solution is to ask them which ports they need, and forward those to them using a port within their assign range. i.e. teach them how to access their home web server using a different port(say 32424, or similar). This won't solve all the issues, which is why we use solution 1. On 27 February 2018 at 09:32, Mike Hammett <nanog@ics-il.net> wrote: I'm a fan of nailing each customer IP to a particular range of ports on a given public IP. Real easy to track who did what and to prevent shifting IPs. ----- Mike Hammett Intelligent Computing Solutions Midwest Internet Exchange The Brothers WISP ----- Original Message ----- From: "Aaron Gould" <aaron1@gvtc.com> To: Nanog@nanog.org Sent: Tuesday, February 27, 2018 10:30:21 AM Subject: cgnat - how do you handle customer issues Couple questions please. When you put thousands of customers behind a cgnat boundary, how do you all handle customer complaints about the following. 1 - for external connectivity to the customers premise devices, not being able to access web servers, web cameras, etc, in their premises? 2 - from the premise natted device, when customers go to a university or bank web site, how do you handle randomly changing ip addresses/ports that may occur due to idle time and session tear-down in nat table such that the bank website has issues with seeing your session ip change? -Aaron
On 02/27/2018 12:52 PM, Aaron Gould wrote:
Thanks
For #2 – what if the ports allocated aren’t enough for the amount of inet traffic the customer site uses ? …is the customer denied service based on insufficient port range ? …or are they assigned another block within that some ip’s range of I think it’s 0-64k or 1025-64k… but how far can you take that before there aren’t anymore port blocks left on that single ip ? …and if you have to allocate that customer another port block from a *different* ip, then we are in the situation of the bank website not liking the fact that the session is bouncing to a different ip maybe ?
Yes, common problem, and the session just fails (TCY SYN dropped) because of insufficient ports. Most CGNs allow you to configure a range of ports for a customer, and reserve an additional range if they need more ports. Yes, if you allocate 1000 ports for one IPv4 address to each of 50 customers and they all need hundreds more ports than that, you're going to run out of ports and connections fail. This is why you have IPv6. Even while many web sites and apps don't support IPv4, enough do that it relieves some pressure on your CGN. Lee
- Aaron
From: Michael Crapse [mailto:michael@wi-fiber.io] Sent: Tuesday, February 27, 2018 11:19 AM To: Mike Hammett Cc: Aaron Gould; NANOG list Subject: Re: cgnat - how do you handle customer issues
For number 2, I'm a fan of what mike suggests. I believe the technical term is MAP-T. For number 1, anyone who wants one, gets one. We provide free public static IP to any customer who asks for one. Another solution, using above solution is to ask them which ports they need, and forward those to them using a port within their assign range. i.e. teach them how to access their home web server using a different port(say 32424, or similar). This won't solve all the issues, which is why we use solution 1.
On 27 February 2018 at 09:32, Mike Hammett <nanog@ics-il.net> wrote:
I'm a fan of nailing each customer IP to a particular range of ports on a given public IP. Real easy to track who did what and to prevent shifting IPs.
----- Mike Hammett Intelligent Computing Solutions
Midwest Internet Exchange
The Brothers WISP
----- Original Message -----
From: "Aaron Gould" <aaron1@gvtc.com> To: Nanog@nanog.org Sent: Tuesday, February 27, 2018 10:30:21 AM Subject: cgnat - how do you handle customer issues
Couple questions please. When you put thousands of customers behind a cgnat boundary, how do you all handle customer complaints about the following.
1 - for external connectivity to the customers premise devices, not being able to access web servers, web cameras, etc, in their premises?
2 - from the premise natted device, when customers go to a university or bank web site, how do you handle randomly changing ip addresses/ports that may occur due to idle time and session tear-down in nat table such that the bank website has issues with seeing your session ip change?
-Aaron
There’s also the issue of what a customer who needs something like GRE or IKE to work does from behind a CGNAT where there aren’t port numbers available for multiplexing. Owen
On Feb 27, 2018, at 2:42 PM, Lee Howard <Lee@asgard.org> wrote:
On 02/27/2018 12:52 PM, Aaron Gould wrote:
Thanks
For #2 – what if the ports allocated aren’t enough for the amount of inet traffic the customer site uses ? …is the customer denied service based on insufficient port range ? …or are they assigned another block within that some ip’s range of I think it’s 0-64k or 1025-64k… but how far can you take that before there aren’t anymore port blocks left on that single ip ? …and if you have to allocate that customer another port block from a *different* ip, then we are in the situation of the bank website not liking the fact that the session is bouncing to a different ip maybe ?
Yes, common problem, and the session just fails (TCY SYN dropped) because of insufficient ports. Most CGNs allow you to configure a range of ports for a customer, and reserve an additional range if they need more ports. Yes, if you allocate 1000 ports for one IPv4 address to each of 50 customers and they all need hundreds more ports than that, you're going to run out of ports and connections fail.
This is why you have IPv6. Even while many web sites and apps don't support IPv4, enough do that it relieves some pressure on your CGN.
Lee
- Aaron
From: Michael Crapse [mailto:michael@wi-fiber.io] Sent: Tuesday, February 27, 2018 11:19 AM To: Mike Hammett Cc: Aaron Gould; NANOG list Subject: Re: cgnat - how do you handle customer issues
For number 2, I'm a fan of what mike suggests. I believe the technical term is MAP-T. For number 1, anyone who wants one, gets one. We provide free public static IP to any customer who asks for one. Another solution, using above solution is to ask them which ports they need, and forward those to them using a port within their assign range. i.e. teach them how to access their home web server using a different port(say 32424, or similar). This won't solve all the issues, which is why we use solution 1.
On 27 February 2018 at 09:32, Mike Hammett <nanog@ics-il.net> wrote:
I'm a fan of nailing each customer IP to a particular range of ports on a given public IP. Real easy to track who did what and to prevent shifting IPs.
----- Mike Hammett Intelligent Computing Solutions
Midwest Internet Exchange
The Brothers WISP
----- Original Message -----
From: "Aaron Gould" <aaron1@gvtc.com> To: Nanog@nanog.org Sent: Tuesday, February 27, 2018 10:30:21 AM Subject: cgnat - how do you handle customer issues
Couple questions please. When you put thousands of customers behind a cgnat boundary, how do you all handle customer complaints about the following.
1 - for external connectivity to the customers premise devices, not being able to access web servers, web cameras, etc, in their premises?
2 - from the premise natted device, when customers go to a university or bank web site, how do you handle randomly changing ip addresses/ports that may occur due to idle time and session tear-down in nat table such that the bank website has issues with seeing your session ip change?
-Aaron
On 02/27/2018 11:30 AM, Aaron Gould wrote:
Couple questions please. When you put thousands of customers behind a cgnat boundary, how do you all handle customer complaints about the following.
1 - for external connectivity to the customers premise devices, not being able to access web servers, web cameras, etc, in their premises?
For web servers, you gently remind them of their ToS, and offer to upgrade them to a business account. You offer to upgrade them to native IPv4, for a fee, and suggest they contact the manufacturer to see why it doesn't support IPv6. I've heard of PCP for this, but never seen it in production on CPE. I saw an early prototype of DS-Lite that included a customer portal that would allow the customer to enable port forwarding, but I can't imagine trying to walk a customer through two layers of port forwarding. Most of these consumer electronics devices long ago started using ICE/TURN/STUN, and/or the company's web portal as their external communication platform. Doesn't mean all of them do.
2 - from the premise natted device, when customers go to a university or bank web site, how do you handle randomly changing ip addresses/ports that may occur due to idle time and session tear-down in nat table such that the bank website has issues with seeing your session ip change?
You can extend your idle timers, of course, but only so much before you're squatting on all your ports. After that, it's down to explaining to the university, bank, etc. that they need IPv6 if they want their web site to be reachable. Or at least they need to extend their idle timers. Lee
-Aaron
participants (6)
-
Aaron Gould -
Chris Gross -
Lee Howard -
Michael Crapse -
Mike Hammett -
Owen DeLong