|
On 1999-05-23T14:22:10,
Joseph Mack <mack@xxxxxxxxxxx> said:
> Just got back and put my computers together at home.
Dito. Jetlag sucks, and lets just say that United Airlines is not HA nor High
Performance...
> There were some people in the audience who were well
> aquainted with the theory. One is on this mailing list
> and hasn't posted anything, so I'll let him remain
> anonymous until he's ready to say something. He's either
> using LVS for a project he has at work or is waiting
> for the 2.2.x version to use in this project.
I gues thats me, so I might as well speak up ;-)
Actually, both: We will start working with the 2.2 patch on a project, since
the 2.0 kernel lacks some other features we need, like serial consoles,
ipchains etc - sure we could patch it, but we would rather try to send
sensible bugreports for the 2.2 patch and not redo the kernel stuff to often
;-)
We have a bunch of stuff to solve for this: syncing configurations between the
load-balancers (since there of course need to be two for HA), which we'll
probably do via a layer on top of heartbeat, syncing the current masquerade
entries from the master to the client (so that sessions don't get lost if a
takeover happens): should be easy if someone does a hook to somehow export
events about entry deletion/creation (via klogd?) and adds a syscall to setup
a new entry. And some further stuff we have not yet thought about I am sure,
but the complete discussion might be suited better for the linux-ha list
instead of here.
We'll give proper credits and contribute all of the code back.
But first, the customer actually needs to sign the order ;-)
> One person was interested in the geographic
> based load director and wanted me to let him
> know when it was done.
I just repeat my comment on the geographic director, maybe we can get some
discussion here: It was proposed that the "tunneling" loadbalancing can be used
for this.
You _really_ do not want to do that via tunneling, since the packets to the
server would still have to go through your loadbalancer at all times, even
though the replies go direct.
You want to extend bind to not only do round robin DNS, but to answer a query
based on the load of the respective clusters and, even though this is slightly
more interesting, proximity to the client. (The F5 3/DNS solution does this)
As soon as we get load-informed load balancing (which is actually easy to do,
as soon as someone implements a command to change the weights on the fly, and
then you just poll for load-average (or some other sensible metric) every x
seconds and set the weight to 100 x loadavg, which should actually work
reasonably well), this is simple: Just do weighted round-robin in the DNS
server and poll the clusters for avg(loadavg) every minute or so.
Sincerely,
Lars Marowsky-Brée
--
Lars Marowsky-Brée
Network Management
teuto.net Netzdienste GmbH - DPN Verbund-Partner
|
