Use service according to priority instead of load

[ct_yau]

Hi,

I want to build a configuration that a CSS11501 can route the client request to servers in a farm according to strict priority, i.e. all request to server A, to server B only when server A fail, to server C only if both A & B fail. The services could be layer 3, 4 or 5. I have read through the documentations and there are only configurations for round-robin, weighted round-robin, or according to load. There is a failover command but it does not seem what I want because it only take effect when some service fail. Is there simple configuration commands to achieve this other than writing customised keepalive scripts?

CT Yau

Hong Kong

[Gilles Dufour · Accepted solution]

you can use a primarysorryserver and secondarysorryserver.

A = normal service

B = primarysorryserver

C = secondarysorryserver

Regars,

Gilles.

[ct_yau]

Thanks Gilles.

But this solution seems not very scalable as only 2 sorry services can be configured. Well, the setup in my original question may not be scalable as well. It's foolish to have 10 servers but use them in strict priority order. Two sorry servers seem reasonable but I just wonder why Cisco impose a limit of 2 in the sorry servers instead of generalizing it (sometime like "add sorryserver B; add sorryserver C;add sorryserver D; ...).

[Gilles Dufour]

The sorryserver was implemented to redirect user to a webserver displaying an error message like 'sorry - site is down'. So the name sorryserver.

That's actually the purpose of the sorryserver.

We didn't do it differently because nobody asked for it.

I don't see the advantage to have servers not being used - just sitting there.

Why don't you want to loadbalance all the traffic over all the available servers ?

Gilles.

[ct_yau]

In my company, we thought of a setup that server A is connected via LAN while some server B is connected via a serial line. (There are some users in the server B location and will access the servers in the reverse order) In that setup, we want server B to provide service only when server A is down. For load balancing according to load, we may for example set the weight for A to 100 and server B to 1, but users still have 1% chance to be connected via a serial line. A strict priority scheme seems a better choice. Why we don't want even the 1% to happen is for both performance and audit trail purpose. Moreover, the servers provide other services in addition to web.

We finally purchased two pairs of CSS with advanced feature set (A CSS with advanced feature set is almost twice the price of one with standard feature set) and implemented zone-based DNS GSLB.

[csbowser]

The Global Site Selector (GSS) models might be able to do what you are asking for, and alot cheaper than buying another pair of enhanced CSS's. Try looking into those to solve the problem.

[ct_yau]

Thanks for the information.

After reading some documentations it seems the GSS is serving the same functions as the enhanced feature set of the CSS. Can I consider a CSS with enhanced feature set as a CSS+GSS? Is there any difference in functionalities?