Backup serverfarm CSS

[claeysg]

Hello,

My customer has two sites, one which is active and one which is backup (disaster recovery). Both sites are interconnected with a 10GB link so we can consider both sites as a single site.

One each site, there are four servers, let's say server1 to 4 on site 1 and server5 to 8 on site 2.

What the customer wants is that by default, all traffic is only load-balanced to server1 to 4. However, if servers1 to 4 go down, then he wants to load-balance the traffic to server5 to 8.

The functionnality that I'm searching for is a kind a backup serverfarm that is only used when all the servers of one serverfarms are down.

Does such functionnality exist on the CSS or do I need the CSM ?

Thanks

[Gilles Dufour]

the CSS only allow 1 backup server.

So, you can achieve what you want with a CSS but not directly.

Since you have 2 sites, you could have a 2nd vip active in the standby site and when the 4 servers are down in the primary site the CSS sends a redirect to the client so they go to the 2nd vip.

Gilles.

[claeysg]

Hello Gilles,

Thanks for your answer.

When you say "So, you can achieve what you want with a CSS but not directly", do you mean that it is not possible to use another vip on the same CSS ?

If I'm right, I have to configure a service type redirect on the primary CSS that points to a VIP on the second CSS. Is it correct ?

Does this work only for http traffic ?

Also, if clients are redirected to the secondary CSS, what will happen with the clients when one of the server goes up again.

Thanks,

Gaetan

[Gilles Dufour]

with http, you can have a redirect that will send the client back to the same css but on a different content rule.

But it only works with HTTP.

This is because we use the http redirect feature to force the client to open a new connection.

For non-http traffic, you can't use an http redirect. So, we keep the connection open but send the traffic somewhere else. So the need for a 2nd CSS that will receive the connection and make a new loadbalancing decision.

When a client is connected to a server, it stays connected to that server even if a new server goes up.

Only new connections from the client would be sent to a different server.

Gilles.

[claeysg]

Hello Gilles,

Thanks for your answer. I think that the application is based on http so perhaps option 1 is a good option.

I have found a configuration example for option 1 on CCO (http://www.cisco.com/en/US/products/hw/contnetw/ps792/products_tech_note09186a00801c65b5.shtml#ex2).

Am I right ?

For option 2 (non http traffic), do you have a configuration example ? How to configure this redirect service ? I do not understand how the client can be redirected. Is the client traffic still passing through the first CSS then ?

Also, with option 2, as I have to redirect the client to a VIP on another CSS, if I want to build a redundant solution (one CSS the backup of the other one with session redundancy between them), is it still possible ? Indeed, the "backup" vip will be defined on both CSS.

Thanks again for your valuable help,

Gaetan

[Gilles Dufour]

For option #1 you have the right config.

For option #2, the traffic will still go to CSS1 which opens a new connection in the back with a remote CSS.

There is no link to a config similar to this one but somebody asked the same question recently and I posted a config for him.

You should be able to find it in the recent forum discussions.

Gilles.

[claeysg]

Hello Gilles,

Thanks again for your answer.

Can you please have a look at the configuration below and give comments ?

As you will see, I added a third option based on vip redundancy. Is this a valid option too ?

-------------------------------------

------------OPTION 1-----------------

---SOLUTION BASED ON HTTP REDIRECT---

-------------------------------------

Two dns entries

http://www.test.com resolves to 1.1.1.1

redirect.test.com resolves to 2.2.2.2

###CSS1###

service redirect

ip address x.x.x.x

keepalive type none

type redirect

domain redirect.test.com

content vip1

vip address 1.1.1.1

add service s1

add service s2

add service s3

add service s4

add service redirect

active

content vip2

vip address 2.2.2.2

add service s5

add service s6

add service s7

add service s8

active

-------------------------------------

------------OPTION 2-----------------

---SOLUTION BASED ON CSS REDIRECT----

-------------------------------------

###CSS1###

service redirect

2.2.2.2

active

content vip1

vip address 1.1.1.1

add service s1

add service s2

add service s3

add service s4

primarysorryserver redirect

active

group redirect

vip address 1.1.1.1

add destination service redirect

active

###CSS2###

content vip2

vip address 2.2.2.2

add service s5

add service s6

add service s7

add service s8

active

-------------------------------------

------------OPTION 3-----------------

---SOLUTION BASED ON VIP REDUNDANCY--

-------------------------------------

###CSS1###

circuit VLAN10

ip address x.x.x.x

ip virtual-router 200 priority 200 preempt

ip redundant-vip 200 1.1.1.1

ip critical-service 200 s1

ip critical-service 200 s2

ip critical-service 200 s3

ip critical-service 200 s4

content vip1

vip address 1.1.1.1

add service s1

add service s2

add service s3

add service s4

active

###CSS2###

circuit VLAN10

ip address y.y.y.y

ip virtual-router 200 priority 100 preempt

ip redundant-vip 200 1.1.1.1

content vip1

vip address 1.1.1.1

add service s5

add service s6

add service s7

add service s8

active

[Gilles Dufour]

about your option #3, it should work.

I have never seen it before so I initially thought it was weird, and tried to find an issue but couldn't.

Gilles.

[claeysg]

Hello Gilles,

Thanks again for your help. I just come back from the customer site where I have presented the different options (option 1 to 3 above) + two additional ones: the use of the csm "backup serverfarm" feature or the ACE blade.

The customer prefers the csm solution or the ACE if this feature is also available.

Can you please check is this feature exists on the ACE ? I checked the configuration guide and have seen the "sorry server" but not the backup serverfarm.

Is this not supported ?

Thanks and have a nice we

[Gilles Dufour · Accepted solution]

the same backup serverfarm option exists in the ACE module

switch/Admin(config)# policy-map type loadbalance first-match SF_FTP_Linux1

switch/Admin(config-pmap-lb)# class class-default

switch/Admin(config-pmap-lb-c)# serverfarm linux1-ftp2 ?

backup Specify backup serverfarm

Carriage return.

switch/Admin(config-pmap-lb-c)# serverfarm linux1-ftp2 backup ?

Enter backup serverfarm name (Max Size - 64)

CACHES

linux1

linux1-ftp2

Gilles.