Re: [tcpm] [v6ops] Flow Label Load Balancing

[Brian E Carpenter <[e-mail]>]

On 25-Nov-20 08:28, Alexander Azimov wrote:
> Hi Fernando,
> 
> Stating that FL change during TCP session lifetime is a bug - is a bit harsh.

It's a bug. It definitely breaks server load balancing, which in practice is more important than en route load balancing, because it will probably switch your session to a different server. (OK, I know that there are mechanisms sometimes for rescuing a session that has unexpectedly moved from one server to another, but in general it's a fail.)

I assume you have read https://tools.ietf.org/html/rfc7098

Regards
    Brian

> 
> It is a fantastic idea to change the FL value if RTO or SYN_RTO happens in a controlled environment. 
> These are very specific TCP timeouts, that provide enough guarantee that there will be no out-of-order packets, though your packets will reach the destination even in case of an outage inside your network. Zero influence on your services in case of the network outage - doesn't sound like a bug for me.
> 
> The problem is with the current Linux (though I haven't checked other OSes) defaults. Specifically, with the default behavior after RTO event.
> 
> вт, 24 нояб. 2020 г. в 22:06, Fernando Gont <[e-mail] <mailto:[e-mail]>>:
> 
>     On 19/11/20 07:48, Alexander Azimov wrote:
>     > Dear colleagues,
>     >
>     > I have added in the cc both v6op and tcpm for a reason and let me
>     > explain why.
>     >
>     > It's clear that we are moving forward with load balancing that uses flow
>     > label (FL). And the pressure will increase with SRv6 adoption. But at
>     > the moment wide adoption of FL-based load-balancing may create
>     > significant issues for TCP Anycast services.
>     >
>     > RFC6437 suggests putting hash from 5-tuple into FL value. And as far as
>     > I know, there is no document that updates this behavior. This
>     > description is perfectly fine, but what is implemented in the Linux
>     > kernel is different: FL is carrying hash from 5-tuple with an additional
>     > seed, and this seed is randomly changed after each RTO/SYN_RTO event.
> 
>     Changing the FL upon RTO is a bug.
> 
>     I guess/assume that when you say SYN-RTO, you really mean "user
>     timeout", rather than RTO. If you don't, then that's also a bug.
>     If you do, I fail to understand what's the reason for wanting the FL to
>     change in that case, because as a result of port randomization, it 0s
>     unlikely that the same four-tuple is employed for the next connection retry.
> 
> 
>     > Here are related patches:
>     >
>     >   * https://lore.kernel.org/netdev/[e-mail]/
>     >   * https://lore.kernel.org/netdev/[e-mail]/
>     >   * https://lore.kernel.org/netdev/[e-mail]/
>     >
>     >
>     > This is a great thing by the way because in the data center
>     > environment with multiple equal paths it gives a way to have
>     > pseudo-multipath TCP which jumps between paths in case of an outage.
>     > There might be interest to writedown an informational document for this.
> 
>     That's a bad idea, since specs-wise the Flow-Label is not guaranteed to
>     remain unchanged from source to destination. If you want to ahve
>     multiple paths, then you should implement that in routing.
> 
> 
> 
>     > I wonder what you think is a proper solution:
>     >
>     >   * Making FL related RTO change as knob instead of default behavior;
>     >   * Adding negotiation behavior in TCP;
>     >   * Something else?
> 
>     Just make the FL a function of the connection "identifier". And keeo it
>     constant for the lifetime of that conenction.
> 
>     Thanks,
>     -- 
>     Fernando Gont
>     e-mail: [e-mail] <mailto:[e-mail]> || [e-mail] <mailto:[e-mail]>
>     PGP Fingerprint: 7809 84F5 322E 45C7 F1C9 3945 96EE A9EF D076 FFF1
> 
> 
> 
> 
> 
> -- 
> Best regards,
> Alexander Azimov
> 
> _______________________________________________
> v6ops mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/v6ops
>