|
Tony Requist wanted us to know:
>
>We currently have a LVS configuration with 2 directors and a set of web
>servers using LVS-DR and keepalived between the directors (and a set of
>MySql servers also). This is all working well using the standard RR
>scheduling without persistence.
>
>We will be adding functionality that will be storing data on some but not
>all web servers. For this, we need to be able to route requests to specific
>web servers according to the URL.
>Ideally I would generate URLs like:
> stuff.domain.com://KEY
>and I could have a little code in LVS (or called from LVS) where I could
>decode KEY to find that the data is on server A, B and C -- then have LVS
>route to one of these three servers.
This is application level, not network level. The better solution
(IMHO) is to put two machines doing reverse proxy with the rules to send
the requests to the correct server. Then have your load balancers
balance among the two rproxies.
A poor man's solution would be to put the reverse proxying on the
webservers themselves.
This is not really good for HA though since you don't have redundancy if
there is only one webserver serving a particular resource.
If the rproxies have a different IP than the public IP of the
webservers, then you have more options.
--
Regards... Todd
They that can give up essential liberty to obtain a little temporary
safety deserve neither liberty nor safety. --Benjamin Franklin
Linux kernel 2.6.11-12mdksmp 1 user, load average: 0.41, 0.24, 0.15
|
