I just have some general question, more or less for pointy haired bosses
rather then myself.
How much is LVS going to be affected by by 2.3, 2.4's multi-threaded TCP
How much is LVS affected in general by the TCP, even if the machines
behind it are 2.2 based machines?
I want to use LVS for a load balancer for the obvious reasons and I
received two emails from the higher ups, first:
Load Balancing under Linux is still in development or hasn't been tested.
We do have 2 Cisco Local Directors (to provide redundancy) in NY that we
can ship to Silicon Valley. Those can be used as load balancers. It will
take about 4 days to configure the local Director with the latest version
and ship them.
I disputed this of course, having used LVS for a while now, I consider it
to be one of the more stable projects, plus the fact that I really don't
know where he's getting this from. It hasn't been tested? Uh, by who, so
I was just confused in general by this.
Next email after my dispute:
I would not use theLinux IP stack and load balancing before kernel 2.4.
Plus, you cannot have a dual system where if one fails the other would
take over without running something like OSPF which will complicate
things. I'll ship you the local directors in the next few days.
So now there's no failover, which of course is false. We all know there's
numerous failover projects that work fine. But the tcp stack of 2.4 is
something I really couldn't comment on (even though I did anyway :-)), so
I want to hear from those who know. I've never had an issue with the tcp
stack with LVS and 2.2, so I just dismiss this as someone somewhere is
against using Linux for this application even though they use it
everywhere else. Whatever, I'm leaving the company by next week, so I
don't care, heh, but I just wanted to know incase these arguments come up