Why associate more than one probe to a serverfarm

Answered Question
May 7th, 2012

Hi,

Can someone describe a case in which is needed to configure a serverfarm with more than one probe associate to it?

Moreover is not clear for me what can happen in case of fails of a single probe. Will it be that all serverfarm will be out fo service in this case? Or all probes associated to the serverfarm have to be failed in order to have the server farm in a out of service state?

I know that in case of more than one probe associated to a real server the ACE is doing a logic AND between them ( fail-on-all command to modify this behavoiur) but not sure if this is still valid for a serverfarm.

Thank you.

I have this problem too.
0 votes
Correct Answer by gaursin2 about 1 year 11 months ago

Hi,

i have mentioned a scenario where in we use multiple probes in single serverfarm (the shopping website e.g).

lets have this config:

serverfarm host SHOPPING

  probe TCP-80

  probe TCP-443

  real server-a

    inservice

  real server-b

    inservice

So ACE will send both TCP-80 and TCP-443 for both servers (A & B).

So if you see show probe detail you wil get two real server bind with each probe. And when you run show rserver you will see status based on probe result and serverfarm confiugration probe logic (AND or OR).

Now if let say TCP-80 probe fails for server A only (taking your q?) then with default mode of AND server-A will be declare dead and rmove from load balancing decision. Same probe is working well for B (along with TCP-443) so that will remian inservice.

Now it will be great if you share your doubt base on above configuration.

  • 1
  • 2
  • 3
  • 4
  • 5
Average Rating: 5 (1 ratings)
gaursin2 Mon, 05/07/2012 - 19:59

Hello,

lets start with probe.

A probe itself has no pass or fail singnificance. Its the real server with which it is binded make it usefull. So a single probe can fail in one instance (when attach to a failed server) and can pass in other instance (when attach to a pass server).

now coming to your q?

A simple exam can be an application which listen for both HTTP and HTTPS (some shopping website, where in via web we choose item and via https we go for monetory transaction). And for such application we create a simple VIP rule of "match vip any", and bind serverfarm to this rule. So its the responsbillity of serverfarm to track health of each server for both TCP-80 and TCP-443. thats why multiple probes are needed.

when this is done the probes are binded to real servers. So in our example, for each server ACE will send TCP-80 and TCP-443 keepalive. Any single probe fail will result in withdrawing that particular server from farm. (logical AND means both probes have to be pass for any server to keep it active). And you are right this default action can be change by using fail-on-all command.

fabio.marino Tue, 05/08/2012 - 01:24

Hi,

your answer is clear and I agree witth you but you are speaking about binding between probe and real server.

My question is related to the scenarion of more probes associated to a serverfarm.

I need  to understand in which case I use this kind of solution and how is the behaviour when all probes associated are in failing state (the instance of course) or for example just one of them is failing state.

Thank you for your answer  anyway.

Bye.

Correct Answer
gaursin2 Tue, 05/08/2012 - 02:19

Hi,

i have mentioned a scenario where in we use multiple probes in single serverfarm (the shopping website e.g).

lets have this config:

serverfarm host SHOPPING

  probe TCP-80

  probe TCP-443

  real server-a

    inservice

  real server-b

    inservice

So ACE will send both TCP-80 and TCP-443 for both servers (A & B).

So if you see show probe detail you wil get two real server bind with each probe. And when you run show rserver you will see status based on probe result and serverfarm confiugration probe logic (AND or OR).

Now if let say TCP-80 probe fails for server A only (taking your q?) then with default mode of AND server-A will be declare dead and rmove from load balancing decision. Same probe is working well for B (along with TCP-443) so that will remian inservice.

Now it will be great if you share your doubt base on above configuration.

fabio.marino Mon, 05/14/2012 - 07:18

Yes I agree with you and I just apply the discussed configuration and the customer is now happy :-)

My last doubt is: do you have any example of scenario that is using a load balancer in 'AND probe-mode' ??

Waiting for your answer and then I will close the discussion. Thank you for ypur time.

Actions

Login or Register to take actions

This Discussion

Posted May 7, 2012 at 2:51 AM
Stats:
Replies:4 Avg. Rating:5
Views:775 Votes:0
Shares:0

Related Content

Discussions Leaderboard

Rank Username Points
1 1,551
2 369
3 333
4 228
5 212
Rank Username Points
5