Application Networking: Keepalive Issue

michaelnolan · ‎09-05-2007

Hi all,

Would anyone have come across a simular issue and maybe help with a root cause as our application team are not finding anything. Is there anything I can do to further prove it's not the CSS.

The service on the CSS for WEB02 checks to see if the asp page(below) is alive, it does this by sending a request to the WEB02 server every 5 seconds checking for the following:

service WEB02

ip address 192.168.11.116

keepalive method get

keepalive type http

keepalive uri "/General/GenAppAlive.asp?ControlAction=InitialLoad"

active

There are a number of retries with a max of three failures before the services is considered to be down on the Load Balancer. Throughout this process the services goes from ALIVE -> DYING to eventually DOWN.

What I see from an ethereal capture:

The WEB02 confirms back to the service on the Load Balancer with a HTTP/1.1 200 ok (text/html) and this confirm that the service is OK. I see at the time the alarm is received that this HTTP/1.1 200 ok (text/html) is missing from the sessions, whereby the packet is not returned from WEB02.

The ASP page connects to ?PCGenAppServerTime.dll? on the App Server as part of the initial load process when called.

Any assistances here much appreciated.

Many thanks,

Regards,

Michael.

Syed Iftekhar Ahmed · ‎09-05-2007

Hi Mike~

It looks as if the page you are probing is a Dynamic page. With CSS if you use GET method in a keepalive, It hashes the page and if the subsequent GET requests gives a different hash the service is marked down.

Try using "keepalive method head" under the service.

Thanks

Syed Iftekhar Ahmed

michaelnolan · ‎09-05-2007

Hi Syed,

I did think of that but I've another WEB server WEB01, using the same asp page and I don't see issues on that.

service WEB01

ip address 192.168.11.115

keepalive method get

keepalive type http

keepalive uri "/General/GenAppAlive.asp?ControlAction=InitialLoad"

active

This is what is throwing me.

Regards,

Michael.

Gilles Dufour · ‎09-10-2007

Michale,

Syed is correct.

If your page is not static, the GET will fail.

Try to use the head method.

If you're not convinced, take a sniffer trace and see what the problem is.

Gilles.

michaelnolan · ‎09-10-2007

Hi Gilles,

But it does not fail all the time, just at a certain period during the night.

Regards,

Michael.

Gilles Dufour · ‎09-10-2007

that's an interesting point.

Do you have to do something to recover ?

If not, it's a question of dynamic page.

A sniffer trace would be required to see what's going on.

Or check if there is anything going on on the server at the time of the failure. Any scripts running ???

The trace would definitely help.

You could also create another service, with the same ip, but different keepalive - like icmp and tcp to see if those go down at the same time.

Gilles.

michaelnolan · ‎09-11-2007

Hi Gilles,

The service recovers by itself. I've checked with the app team and they say the ASP page is a static one.

With the sniifer what I see is that the HTTP/1.1 200 ok is missing, which I know is what causes the service to enter down state. I've placed the sniifer and the CSS to identify this. I cannot see anything else happenning between the WEB server and the CSS.Maybe I need to place the sniifer between the Application server and the Web server to see what is happenning here?

Thanks for your help to date

Regards,

Michael.

Gilles Dufour · ‎09-11-2007

Michael,

I would say the css correctly detects there is an issue with the server.

Now, you'll have to track down what is going on with the server.

As you said, you can continue with the sniffer and see if there is a problem with the application.

Gilles.