search

socallinuxexpo / scale-network

SCaLE's on-site expo network configurations, wifi, tooling, and scripts
https://www.socallinuxexpo.org/
BSD 3-Clause "New" or "Revised" License
40 stars 16 forks source link

Incoporate notification of rate limiting in conference checklist #751

Open sarcasticadmin opened 3 months ago

sarcasticadmin commented 3 months ago

Description

Update docs/conference checklist: https://docs.google.com/document/d/1bIdJmbOmKFKKf7YQjrVjy0aw3g8ajMC8krswPXb-GyM/edit?usp=sharing

The goal is to lessen the potential of attendees being rate limited while at the conference.

The following potential rate limit providers that should be contacted prior to the conference starting with our public IPv4 address and IPv6 subnet:

  1. GitHub
  2. Docker
  3. Fastly (they front cache.nixpkgs.org and seemed to slow us down at one point in both the conference center and hotel)

Additionally it would be a good idea to have user workarounds: Auth Token, etc.

Acceptance Criteria

owendelong commented 3 months ago

We never get the public IPv4 address before Monday.

IPv6 is a non-issue as they don’t seem to rate-limit on the /48, but on the /128.

Since every host gets an individual IPv6 address, there’s a unique /128 per device.

Owen

On Mar 22, 2024, at 12:48, Robert James Hernandez @.***> wrote:

Description

Update docs/conference checklist: https://docs.google.com/document/d/1bIdJmbOmKFKKf7YQjrVjy0aw3g8ajMC8krswPXb-GyM/edit?usp=sharing

The goal is to lessen the potential of attendees being rate limited while at the conference.

The following potential rate limit providers that should be contacted prior to the conference starting with our public IPv4 address and IPv6 subnet:

GitHub Docker Fastly (they front cache.nixpkgs.org and seemed to slow us down at one point in both the conference center and hotel) Additionally it would be a good idea to have user workarounds: Auth Token, etc.

Acceptance Criteria

Documentation in conf checklist for providers that could rate limit attendees Documentation for workarounds to rate limits imposed by providers — Reply to this email directly, view it on GitHub https://github.com/socallinuxexpo/scale-network/issues/751, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAK6GTWUKPV424IQPFA2JJDYZSDHNAVCNFSM6AAAAABFD4TPRCVHI2DSMVQWIX3LMV43ASLTON2WKOZSGIYDGMRQGIZTINA. You are receiving this because you are subscribed to this thread.

sarcasticadmin commented 3 months ago

We never get the public IPv4 address before Monday.

Yep thats why its in the conference instead of preconf checklist

IPv6 is a non-issue as they don’t seem to rate-limit on the /48, but on the /128.

Thats good to know, I figured we could cover our basis on ipv6 just so each provider has that information.

genebean commented 3 months ago

We never get the public IPv4 address before Monday.

Do you know why this is @owendelong?

sarcasticadmin commented 3 months ago

Do you know why this is @owendelong?

PCC and company allocate the ipv4 address off their existing router when we show up on monday. Even if we got it ahead of time I wouldnt surprise me that its subject to change

genebean commented 3 months ago

Seems odd they wouldn't know that address well in advance as I seriously doubt it changes... maybe I'm over thinking it though

owendelong commented 3 months ago

It does. They’ve changed ISPs at least 3 times in the time we’ve been holding our event there. Prior to this year, we were getting a /29 from them. This year, we got a single /32.

It appears that they have reduced their overall range from it’s previous /26 to now being just a /29 for the entire PCC.

I won’t defend or even attempt to explain their conduct, but Encore does what Encore does and we just have to cope with it as best we can.

On Mar 23, 2024, at 18:41, Gene Liverman @.***> wrote:

Seems odd they wouldn't know that address well in advance as I seriously doubt it changes... maybe I'm over thinking it though

— Reply to this email directly, view it on GitHub https://github.com/socallinuxexpo/scale-network/issues/751#issuecomment-2016658752, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAK6GTTFUZLFEDHM2ACLNPLYZYVMXAVCNFSM6AAAAABFD4TPRCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDAMJWGY2TQNZVGI. You are receiving this because you were mentioned.

irabinovitch commented 3 months ago

We reduced to a /32 because we said we didn't need more than 1 IP and it allowed significant savings.

MrHamel commented 3 months ago

Is the significant savings worth the blowback from the rate limiting? Owen is against running a NAT pool on the firewalls, but it could make it less of an issue, if for example we had two IPs (one for each building).

@sarcasticadmin I have a few contacts at Fastly that I can poke, to see if they're willing to make some exceptions, can't make any promises though.

irabinovitch commented 3 months ago

Given we paid many many thousands of dollars per year for 5+ years for the extra IPs and never used them, yes its worth the savings.

davidelang commented 3 months ago

On Thu, 28 Mar 2024, Ryan Hamel wrote:

Is the significant savings worth the blowback from the rate limiting? Owen is against running a NAT pool on the firewalls, but it could make it less of an issue, if for example we had two IPs (one for each building).

I would prefer not to break sessions when people move from one building to the other.

David Lang

sarcasticadmin commented 3 months ago

Is the significant savings worth the blowback from the rate limiting? Owen is against running a NAT pool on the firewalls, but it could make it less of an issue, if for example we had two IPs (one for each building).

@MrHamel considering we got rate limited in < 1hr on Thursday I dont see there being much benefit to having a few more IPs since it doesnt by us much additional overhead for the number of requests.

@sarcasticadmin I have a few contacts at Fastly that I can poke, to see if they're willing to make some exceptions, can't make any promises though.

:+1: thatd be much appreciated

sarcasticadmin commented 3 months ago

I would prefer not to break sessions when people move from one building to the other.

@davidelang the address space between the buildings for wireless is already different: https://github.com/socallinuxexpo/scale-network/blob/c6e657d6cca7fc17aac370905dd1e8105d087f8b/switch-configuration/config/vlans.d/Expo#L2-L3

https://github.com/socallinuxexpo/scale-network/blob/c6e657d6cca7fc17aac370905dd1e8105d087f8b/switch-configuration/config/vlans.d/Conference#L2-L3

davidelang commented 3 months ago

On Fri, 29 Mar 2024, Robert James Hernandez wrote:

I would prefer not to break sessions when people move from one building to the other.

@davidelang the address space between the buildings for wireless is already different: https://github.com/socallinuxexpo/scale-network/blob/c6e657d6cca7fc17aac370905dd1e8105d087f8b/switch-configuration/config/vlans.d/Expo#L2-L3

yep, realized that just after I sent the message, sorry.

David Lang

owendelong commented 3 months ago

Actually, we did use them. We made changes this year to reduce our IPv4 utilization and we also had used them previously some years for an Akamai cache server. We don’t need them any more,  but it is not far to say we did not use them. This year is also the first time you mentioned that the multiple IPv4 addresses were costing us extra. OwenOn Mar 29, 2024, at 05:32, Ilan Rabinovitch @.***> wrote: Given we paid many many thousands of dollars per year for 5+ years for the extra IPs and never used them, yes its worth the savings.

—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you were mentioned.Message ID: @.***>

owendelong commented 3 months ago

On Mar 29, 2024, at 09:57, David Lang @.***> wrote: On Thu, 28 Mar 2024, Ryan Hamel wrote:

Is the significant savings worth the blowback from the rate limiting? Owen is against running a NAT pool on the firewalls, but it could make it less of an issue, if for example we had two IPs (one for each building).

I would prefer not to break sessions when people move from one building to the other. They break anyway because their internal address will change. I don’t see a significant gain to doing separate addresses per building because the rate limiting was mostly workshop related which meant it was mostly in one building anyway.  David Lang

—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you were mentioned.Message ID: @.***>