From: Ramcharan, Vijay A (vijay.ramcharan@verizonbusiness.com)
Date: Tue Dec 02 2008 - 13:39:31 ARST
I don't believe that you have defined your problem and the desired goal
appropriately.
You stated that you "need a fallback to default round robin in case
server becomes unavailable".
So one of your real servers fails which is a member of a serverfarm.
Naturally since this is a sticky session I assume that you have a reason
for it to be sticky (shopping cart for example).
Therefore you may want to have the client/user reconnect to a different
real server and start all over again with whatever they were doing when
the real server failed.
In that case, I would look into the "failaction purge" command.
Options:
a) Purge existing connections going to that real server
b) Let the LB make the decision which by your own admission appears to
be that it sends the connections over to the next available server(s)
c) ...fill in what other options may be applicable in your situation
Really though, I think you need to determine what the LB "should" do in
your server failure scenario and then find the appropriate knobs to turn
to achieve the desired goal.
Vijay Ramcharan
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
Lora Ganeva
Sent: December 02, 2008 04:00
To: Thomas Fowles
Cc: ccielab@groupstudy.com
Subject: RE: ACE with sticky serverfarm
Hi Thomas,
Thanks a lot. I am pretty sure that my question will be not answer in
the
official cisco site. I have configured the stickness, it is working but
in an
unpredictable manner - when server fails, connections (that have to be
sticky)
are for some reason forwarded to the next server in the serverfarm.
BR<
Lora
From: Thomas Fowles [mailto:tfowles@gmail.com]
Sent: 02 dEKEMWRI 2008 G. 02:53
To: Lora Ganeva
Cc: ccielab@groupstudy.com
Subject: Re: ACE with sticky serverfarm
This link might be helpful to you:
http://www.cisco.com/en/US/docs/app_ntwk_services/data_center_app_servic
es/ac
e_appliances/vA1_7_/configuration/slb/guide/classlb.html#wpxref39283
Basically, you configure two serverfarms, a primary and a backup. When
all
servers in the primary are gone, you fail back to the sorry serverfarm,
and it
looks like you can specify a different predictor (lb-algorithm) for each
serverfarm if you wish (ie leastconn & round-robin). I'm not sure that
you
can make the primary sticky and not the backup.
HTH,
Thomas Fowles
CCIE#18762
On Mon, Dec 1, 2008 at 6:28 AM, Lora Ganeva <lganeva@mobiltel.bg> wrote:
Hi experts,
If anyone has in depth experience with ACE please let me know. I am
facing a big problem with sticky serverfarm that need a fallback to
default round robin in case server becomes unavailable.
10x in advance,
Lora
Blogs and organic groups at http://www.ccie.net
This archive was generated by hypermail 2.1.4 : Thu Jan 01 2009 - 12:53:07 ARST