Cisco ACE "sticky" question

[rathinilesh]

Hi,

I have a question about how a static entry under a "sticky" performs

Configuring Static IP Address Sticky Table Entries

Cisco Documentation Says When you configure a static entry, the ACE enters it into the sticky table immediately.

Configuring the ACE Action on Server Failure

failaction purge

The purge keyword specifies that the ACE remove the  connections to a real server  if that real server in the server farm  fails after you enter the  command. The ACE sends a reset (RST) to both  the client and the server  that failed.

Cisco Documentation Says If you do not configure this command, the ACE takes no action when a server fails

sample config

sticky ip-netmask 255.255.255.240 address source STICKY1
   timeout 180
   replicate sticky
   serverfarm SERVERFARM1
   8 static client source 192.168.12.15 rserver SERVER1

Question1 - What happens if SERVER1 fails?

a) Does the ACE let the connections to SERVER1 timeout(default behaviour) and then load-balance new connections coming in deom 192.168.12.15 to another server in SERVERFARM1

OR

b)  Does the ACE reset the connections to SERVER1  immediately and starts  load-balancing new conenction coming in from  192.168.12.15 to other  servers in SERVERFARM1 ?

OR

c) Does the ACE just drop the current and new connections from 192.168.12.15 till SERVER1 comes back up ?

OR

d) Is it dealt differently?

Question2 - Now what happens if the failed server(SERVER1) comes back up after some time?

e) Does the ACE reset any current connections from 192.168.1.15 and starts sending them to SERVER1 ?

OR

f)  Does the ACE leave the current connections from 192.168.1.15 to other  servers in SERVERFARM1 as they are and send any new connections

from 192.168.1.15 to SERVER1?

OR

g) Is it dealt differently?

My guess is Question1 -> a) and Question2 -> e)

ACE model =  ACE10-6500-K9

Version =  A2(3.3)

Any help on this issue would be much appreciated.

Regards,

Nilesh

[Ahmad Basel Jaber · Accepted solution]

Hi Nilesh,

The decision was taken by BU that if the servers which has static sticky entries point to fails, and therefore are not reachable, the static sticky entries will not be bypassed and connections are continuously sent to the 'out-of-service' servers.

If you need redundancy with static sticky entries, then you should use backup real server within a servefarm:

http://www.cisco.com/en/US/docs/interfaces_modules/services_modules/ace/v3.00_A2/configura
tion/slb/guide/rsfarms.html#wp1000264

Best regards,

Ahmad

[rathinilesh]

Hi Ahmad,

thank you for your reply. One last question.

We can configure a resource-limit per context including that for sticky entries.

Is there limit or a way to limit the number of sticky entries per sticky table within an ACE context ?

Regards,

Nilesh

[Ahmad Basel Jaber]

Hi Nilesh,

No, the only control level you have is per context bases, then the resource will be shared between all sticky groups/tables inside the context.

Best regards,

Ahmad

[Ahmad Basel Jaber · Accepted solution]

Hi Nilesh,

I have done more tests in my lab regarding this issue, and I found my information was wrong

Please refer to the following document which should cover all scenarios:

https://supportforums.cisco.com/docs/DOC-17765

Best regards,

Ahmad