[f-nsp] virtual host-range, dsr, and health checks, oh my!

[Matt Stockdale]

Gentlepeople -

  Recently I've tried the "host-range" method of load-balancing a large
(36) number of consecutive VIP's to 2 real web servers using DSR. At
first, it "worked" (at the time, only the first 2 of the 36 VIP's were
bound as loopbacks to each of the real servers), but I've run into some
behavior that leads me to question the viability of this method not only
for my needs, but under any circumstance. Namely-

  1. Even if the application ports on the real servers (http and ssl, in
this case) are "Failed", the VIPs are still directing traffic to the
real server(s).

  2. Only the first VIP in the range seems to be receiving a layer-4
health check. Furthermore, even when it fails, other VIPs seem
unaffected. 

  3. (maybe more 2a) If, for example, the first real server does _not_
have a particular VIP bound as loopback, but the second real server
does, the serveriron is sending the first request to the unconfigured
first server every time (since I'm still testing, there is no traffic to
the VIPs at this point)

 I'm using ServerIron XL's running 7.4.0T12 code.

  Does anyone use the host-range feature in any sort of serious
deployment? Are there workarounds for the issues I've run into?

  Thanks in advance,
   

  Matt


--