(Sorry for repost. Chris kindly informed me that my messages are empty. I think it could be because GPG sig. Another attempt without it)
Hi,
I am aware there will be a slim chance for this to happen but having a 44net subnet directly routed would be super amazing. I am aware that "normal" providers likely won't do this except for business customers (at many $$$/months) and likely the subnet needs to be large (>=/24?)
I could imagine though that small providers may be up for it (as an example, in a place I lived many years ago, I had a small provider; I contacted the CEO directly and he immideately agreed to assign+route me a /29 at no cost!).
Are there any known providers (US, California) who could route a 44net subnet?
To everyone who has their 44 routed directly: How does it work for you?
Finally: The whole 44net is announed only by the UCSD gateway (as far as I understand). Wouldn't it be great to improve the connectivity, reliability, redundancy by having 44 announced by multiple people who route their subnet directly and then forwarding them to the mesh network via ipip? Is there a reason this is not done, other than nobody besides UCSD volunteered ?
Thanks, KM6RDV
Hi Nick,
I am aware there will be a slim chance for this to happen but having a 44net subnet directly routed would be super amazing. I am aware that "normal" providers likely won't do this except for business customers (at many $$$/months) and likely the subnet needs to be large (>=/24?)
We do allow members to announce their own prefix via BGP, some have their own ASN and do it themselves, others have an existing provider do it for them. A popular choice is Vultr who will rent you a cheap VM and you can run Bird/FRR/Quagga/etc on it and announce your prefix to them via a private ASN, they will then announce it to the DFZ under their own ASN. You can then either use the address space on the VM, or tunnel back to your shack for use there.
Relating to your other question on the list - we do delegate reverse DNS for BGP prefixes.
I could imagine though that small providers may be up for it (as an example, in a place I lived many years ago, I had a small provider; I contacted the CEO directly and he immideately agreed to assign+route me a /29 at no cost!).
Are there any known providers (US, California) who could route a 44net subnet?
To everyone who has their 44 routed directly: How does it work for you?
See above.
Finally: The whole 44net is announed only by the UCSD gateway (as far as I understand). Wouldn't it be great to improve the connectivity, reliability, redundancy by having 44 announced by multiple people who route their subnet directly and then forwarding them to the mesh network via ipip? Is there a reason this is not done, other than nobody besides UCSD volunteered ?
Announcing a prefix yourself and joining the IPIP mesh are mutually exclusive. The TAC are working on creating POPs that will provide additional connectivity options, so fairly soon (I hope) you should be able to connect via a simple VPN, or BGP using a smaller than /24 prefix, etc.
Regards, Chris
Thanks, KM6RDV
44Net mailing list 44Net@mailman.ampr.org https://mailman.ampr.org/mailman/listinfo/44net
On 26 Aug 2021, at 08:35, Rob PE1CHL via 44Net 44net@mailman.ampr.org wrote:
On 8/26/21 9:33 AM, Chris Smith via 44Net wrote:
Announcing a prefix yourself and joining the IPIP mesh are mutually exclusive.
Can you explain?
I mean that members either request a /24 (or larger) and announce it via BGP for their project, or they (typically) request a smaller prefix and join the IPIP mesh / AMPRNet.
Chris
Rob _________________________________________ 44Net mailing list 44Net@mailman.ampr.org https://mailman.ampr.org/mailman/listinfo/44net
On 8/26/21 9:42 AM, Chris Smith via 44Net wrote:
On 26 Aug 2021, at 08:35, Rob PE1CHL via 44Net 44net@mailman.ampr.org wrote:
On 8/26/21 9:33 AM, Chris Smith via 44Net wrote:
Announcing a prefix yourself and joining the IPIP mesh are mutually exclusive.
Can you explain?
I mean that members either request a /24 (or larger) and announce it via BGP for their project, or they (typically) request a smaller prefix and join the IPIP mesh / AMPRNet.
Chris
But you can have a /24 or larger subnet announced on BGP and have the same subnet on the IPIP mesh as well. In fact this improves connectivity for hosts within AMPRnet that are not routing towards internet or do so with limitations (e.g. NAT).
When the IPIP mesh has been replaced with the new backbone, it would be the default situation to be on that backbone network and have that announce your subnet on BGP, as I see it. So it would still mean you can route from/towards internet via your BGP announcement, and towards other AMPRnet users via the backbone.
Rob
On 26 Aug 2021, at 09:03, Rob PE1CHL via 44Net 44net@mailman.ampr.org wrote:
On 8/26/21 9:42 AM, Chris Smith via 44Net wrote:
On 26 Aug 2021, at 08:35, Rob PE1CHL via 44Net 44net@mailman.ampr.org wrote:
On 8/26/21 9:33 AM, Chris Smith via 44Net wrote:
Announcing a prefix yourself and joining the IPIP mesh are mutually exclusive.
Can you explain?
I mean that members either request a /24 (or larger) and announce it via BGP for their project, or they (typically) request a smaller prefix and join the IPIP mesh / AMPRNet.
Chris
But you can have a /24 or larger subnet announced on BGP and have the same subnet on the IPIP mesh as well. In fact this improves connectivity for hosts within AMPRnet that are not routing towards internet or do so with limitations (e.g. NAT).
This has been tried, but the routing setup gets very complicated and it is easy to get it wrong, a way back Brian got me to put code in the portal that disallows this after a few occasions when folk tried to do this and used their BGP announced 44 address as the IPIP gateway IP - it broke things badly!
When the IPIP mesh has been replaced with the new backbone, it would be the default situation to be on that backbone network and have that announce your subnet on BGP, as I see it. So it would still mean you can route from/towards internet via your BGP announcement, and towards other AMPRnet users via the backbone.
Yes, I agree. With a better replacement for the IPIP mesh that the backbone/POP system will bring, this will be possible and even preferred.
Chris
Rob _________________________________________ 44Net mailing list 44Net@mailman.ampr.org https://mailman.ampr.org/mailman/listinfo/44net
On 8/26/21 10:10 AM, Chris Smith via 44Net wrote:
But you can have a /24 or larger subnet announced on BGP and have the same subnet on the IPIP mesh as well. In fact this improves connectivity for hosts within AMPRnet that are not routing towards internet or do so with limitations (e.g. NAT).
This has been tried, but the routing setup gets very complicated and it is easy to get it wrong, a way back Brian got me to put code in the portal that disallows this after a few occasions when folk tried to do this and used their BGP announced 44 address as the IPIP gateway IP - it broke things badly!
That is right, but that is a different problem. A gateway like this should have both an external IP outside of net-44 to be used as the tunnel endpoint, and the subnet it announces on BGP and IPIP which is within net-44. In that case it works very well. Of course you need the proper routing setup, and the proper source address selection for outgoing connections. It is always easiest when "the router" (an actual router, or a general-purpose machine that does the routing) does not host any services itself. (so the source address is already determined on another machine and the traffic is just passed on) But it is possible to make it work OK even when services are hosted, with a little care (and testing).
It is important to have all routes in a single routing table on that router so it considers all routes in subnet-size order (smallest subnet first) and will find the IPIP route for gateways before it considers routing to internet. Solutions where different types of routes are put in different tables that are then used with policy routing do not work well for this case. (they are OK and preferable for stations that are only on the IPIP mesh, as described in the wiki)
Rob
On 26/08/2021 11:10, Chris Smith via 44Net wrote:
But you can have a /24 or larger subnet announced on BGP and have the same subnet on the IPIP mesh as well. In fact this improves connectivity for hosts within AMPRnet that are not routing towards internet or do so with limitations (e.g. NAT).
This has been tried, but the routing setup gets very complicated and it is easy to get it wrong, a way back Brian got me to put code in the portal that disallows this after a few occasions when folk tried to do this and used their BGP announced 44 address as the IPIP gateway IP - it broke things badly!
Actually, this broke thing since people insisted to use one of the subnet's address as the GW. Using an extra public IP for the gateway itself works flawless. Also another issue is the default tunneling of not-announced IPs (all the 44 space) by default via AMPR-GW which has become standard practice, but breaks things. Anyway, the current rip daemons and scripts support this type of operation, too (but it seems to be buggy since it was never extensively long term tested).
At the moment there is 5 systems configured like that: 44.94.17.128/27 and 44.130.104.0/24, 44.130.105.0/24, 44.130.106.0/24, 44.130.107.0/24
I don't know about the first one, but the last four are the system used for testing the daemons and script and are not active anymore (it worked as expected at that time).
73's! Marius, YO2LOJ
Hi Chris,
On 2021-08-26 00:33, Chris Smith wrote:
Hi Nick,
I am aware there will be a slim chance for this to happen but having a 44net subnet directly routed would be super amazing. I am aware that "normal" providers likely won't do this except for business customers (at many $$$/months) and likely the subnet needs to be large (>=/24?)
We do allow members to announce their own prefix via BGP, some have their own ASN and do it themselves, others have an existing provider do it for them. A popular choice is Vultr who will rent you a cheap VM and you can run Bird/FRR/Quagga/etc on it and announce your prefix to them via a private ASN, they will then announce it to the DFZ under their own ASN. You can then either use the address space on the VM, or tunnel back to your shack for use there.
Wow, this is amazing! Thank you very much for this info.
Initially I thought I wanted an *ISP* who routes me the subnet but now that I think about it, a solution like Vultr seems to be the better choice because it's ISP independent.
It seems there is not even a charge to announce a net over BGP.
-> Does anyone who uses Vultr already want to provide me with a referral? (PM)
-> Out of curiosity, are there also people out there who managed to get their 44 routed via their ISP?
Best, Nick
On Aug 28, 2021, at 04:37, Nick KM6RDV via 44Net 44net@mailman.ampr.org wrote: Wow, this is amazing! Thank you very much for this info.
Initially I thought I wanted an *ISP* who routes me the subnet but now that I think about it, a solution like Vultr seems to be the better choice because it's ISP independent.
It seems there is not even a charge to announce a net over BGP.
-> Does anyone who uses Vultr already want to provide me with a referral? (PM)
-> Out of curiosity, are there also people out there who managed to get their 44 routed via their ISP?
Best, Nick
I have had a /23 allocation to a local (to me) data center for over eight years. I have a server there, that I own, on a rented shelf in a shared cabinet. It was a major PIA for them to set it up, requiring negotiation with all of their upstream providers. Took them weeks to get it going, but ultimately successful, all for no extra charge.
I also manage a /20 allocation to Vultr for IRLP.net (used to deliver public IPs to individual repeaters, buried behind NAT). With Vultr.com, the process is entirely automated, fill out a form, apply for a private ASN, Vultr exchanges email with Chris. Then fully working within a few hours.
What is nice about Vultr is the flexibility to move things around. Once your allocation is established with them, you can bring up a /24 (or larger) subnet from any of their 17 data centers around the world, with no notification to anyone. You must configure your host to run a BGP routing daemon. Documentation on how to set that up is on their site. Not difficult. If you check, you will find hundreds of 44.0 subnets routed by Vultr (AS20473). They have been providing that service for many years.
Since most of IRLP is in North America and Australia, we are currently providing services from Chicago and Sydney. $6/mo each. We expect to add Seattle soon, and Dallas after that.
I am not aware of any residential retail ISP that offers BGP routing services. Technically they could, but is simply something they are not inclined to do.
— Dave K9DC
Hi Chris,
On 2021-08-26 00:33, Chris Smith wrote:
Hi Nick,
I am aware there will be a slim chance for this to happen but having a 44net subnet directly routed would be super amazing. I am aware that "normal" providers likely won't do this except for business customers (at many $$$/months) and likely the subnet needs to be large (>=/24?)
We do allow members to announce their own prefix via BGP, some have their own ASN and do it themselves, others have an existing provider do it for them. A popular choice is Vultr who will rent you a cheap VM and you can run Bird/FRR/Quagga/etc on it and announce your prefix to them via a private ASN, they will then announce it to the DFZ under their own ASN. You can then either use the address space on the VM, or tunnel back to your shack for use there.
44Net mailing list 44Net@mailman.ampr.org https://mailman.ampr.org/mailman/listinfo/44net
On Sat, Aug 28, 2021 at 7:35 AM, Dave Gingrich 44net@mailman.ampr.org wrote:
I am not aware of any residential retail ISP that offers BGP routing services. Technically they could, but is simply something they are not inclined to do.
Small local ISPs can and do route to end-users. I had my 44net allocation set up that way with monkeybrains.net until I moved stuff into a datacenter - they announced it for me for a while, then peered with me at my home. But they're pretty unusual.
-Paul
-
Chris Smith -
Dave Gingrich -
Marius Petrescu -
Nick Hammler -
Nick KM6RDV
-
Paul McMillan -
Rob PE1CHL