|
Benjamin Wall wrote:
> The loadbalancing is working on the 2 webservers, I have not set it up
> on the database servers yet because of concerns about database
> connection state and how LVS handles that (or doesn't).
The single writer, many reader problem has not been solved in
the general sense for distributed computing. It's difficult.
LVS can't do it, databases (unless you're willing to shell
out lots of money) don't do it easily. In these situations,
something has to pass the write to all machines.
> What I want:
>
> Two Virtual servers (Direct Routing):
> - 1 for http/apache (loadbalanced) (2 machines, more later)
> - 1 for MySQL (simple failover) (2 machines)
> Automatic addition/removal from LVS when failure detected
> Director Failover (I have two machines for director duty)
>
> What I am confused about:
To a newbie it is confusing. I've been on this project for 5yrs or so
now, and I have a hard time keeping track of LVS and everything
that hangs off it.
there are several software packages with similar functionality.
Any of them will do the job. Mon was used initially, but
no-one seems to be using it now. The maintainers of the other
packages are all on this mailing list and can answer questions.
You're going to have to figure out which one you want to use
and then start installing. I know figuring it out is a bit
of work, but that's what we have.
> Director failover:
> - I am planning on setting up heartbeat via serial cable.
> - Which tool should I use for the takeover? vrrpd, Fake
> - ldirectord, keepalived ??? or just mon, hearbeat, fake?
> - Do I need connection table synchronization?
> http://marc.theaimsgroup.com/?l=linux-virtual-server&m=105459391703228&w=2
we didn't have the sync demon till relatively recently. If you don't mind users
loosing connection on a failure, you don't need any of this.
If you're providing state of the art service, then you need everything.
I would get things going in stages, not to overwhelm yourself.
Are the databases contacted via the webservers, or is the user's
connection to the databases separate to the webservers? You
will have to figure out the single writer, many reader
problem for the databases first, then figure out failover
for the realservers, then failover for the directors.
> - iproute2's advantage over eth0:10 style (is it just an iptables
> issue?) ( I know next to nothing about iproute2 )
as of 2.4 kernels, any new networking functionality requires the
iproute2 style of setting up nics. iptables knows nothing about eth0:10
style aliases. As long as you only have one IP on a NIC you don't need
iproute2.
Joe
--
Joseph Mack PhD, High Performance Computing & Scientific Visualization
SAIC, Supporting the EPA Research Triangle Park, NC 919-541-0007
Federal Contact - John B. Smith 919-541-1087 - smith.johnb@xxxxxxx
|
