Difference between "Connections"

b.petronio · ‎02-08-2008

Hi all,

I've noted a difference between 2 services, which uses Oracle WebCache each one.

One have the field "Total Reused Conns: 0", and the other have the same field with hits.

The counters have been cleared before the 2 hours stress test.

Any ideas why is this happening and what should be better ? zero reused connections, or max reused connections?

Best Regards,

Bruno PetrÃ³nio

service esl0011-7778

ip address 10.1.1.120

port 7778

keepalive type http

keepalive port 7778

keepalive uri "/"

active

service esl0012-7778

ip address 10.1.1.121

port 7778

keepalive type http

keepalive port 7778

keepalive uri "/"

active

content HTTP-teste-tempaplj

vip address 10.1.2.136

redundancy-l4-stateless

port 80

protocol tcp

advanced-balance arrowpoint-cookie

arrowpoint-cookie browser-expire

add service esl0011-7778

add service esl0012-7778

active

CSS_IFAP_2# show service esl0011-7778

Name: esl0011-7778 Index: 5

Type: Local State: Alive

Rule ( 10.1.1.120 ANY 7778 )

Session Redundancy: Disabled

Redirect Domain:

Redirect String:

Keepalive: (HTTP-7778:HEAD:/ 5 3 5 )

Keepalive Encryption: Disabled

Last Clearing of Stats Counters: 07/02/2008 16:48:01

Mtu: 1500 State Transitions: 2

Total Local Connections: 24228 Total Backup Connections: 0

Current Local Connections: 0 Current Backup Connections: 0

Total Connections: 24228 Max Connections: 65534

Total Reused Conns: 0 Weight Reporting: None

Weight: 1 Load: 2

CSS_IFAP_2# show service esl0012-7778

Name: esl0012-7778 Index: 8

Type: Local State: Alive

Rule ( 10.1.1.121 ANY 7778 )

Session Redundancy: Disabled

Redirect Domain:

Redirect String:

Keepalive: (HTTP-7778:HEAD:/ 5 3 5 )

Keepalive Encryption: Disabled

Last Clearing of Stats Counters: 07/02/2008 16:48:01

Mtu: 1500 State Transitions: 0

Total Local Connections: 23494 Total Backup Connections: 0

Current Local Connections: 0 Current Backup Connections: 0

Total Connections: 23494 Max Connections: 65534

Total Reused Conns: 36266 Weight Reporting: None

Weight: 1 Load: 2

Gilles Dufour · ‎02-08-2008

You reuse the connection when the cookie points to the already open connection.

It's not a bad thing that you don't reuse it.

Moreover, you do not have the command 'no persistent', so anyway connections should stay with the same server all the time.

Gilles.

b.petronio · ‎02-08-2008

Thanks for your quick clarification.

But, ( i've always a miserable BUT ), shouldn't be the both services with the same aspect in the "Total reused connections" field, since the both OAS w/ Webcache modules have equal configuration, and the content rule, says to stick the connection by the browser ?

Or am i messing all up, in the balance and sticky options ?

What i want is balance by browser and sticky the connection by browser too, wich i thought the arrowpoint-cookie parameters do it.

Best Regards,

Bruno PetrÃ³nio

Gilles Dufour · ‎02-08-2008

Bruno,

I don't see why one would increment and not the other.

But, that counter is not an indication that stickyness is working or not.

Because you did not configure 'no persistent' the counter should not increment actually.

Do you have sticky connection problem ?

If there is no other concern than the counter itself, we can try to sniff the traffic, see when it increases exactly and when it does not see and from there decide if there is a problem or not.

If the problem is more than the counter - we'll need to also take a sniffer trace of the connection that fails to stick.

Gilles.

b.petronio · ‎02-08-2008

In fact i don't see the any problem in the test's we are performing, but since i'm capturing results to find a best solution for the client, i realize that counter mismatch info.

We are performing stress test's, and the capture is not easy to see, since, we have 3 pc's simulating 75 simultaneous clients in each pc.

We have tried another solution w/ an apache server on the OAS server, and was not showing any reused connections too, but no errors was reported in the log test application, so i assume that the connections were always stickned by the arrowpoint-cookie browser expire parameter.

I'm making this workaround tests to perform compression apart off the CSS, cause it was appearing some errors in that log application that saying connection reset's, when the compression was made by the CSS.

I know that that is not the same of real browser's, but the client is making analysis based on that JMeter app.

There are any reliable application to perform stress test to the CSS ?

Best Regards,

Bruno PetrÃ³nio

Gilles Dufour · ‎02-08-2008

Bruno,

You can try to reduce the traffic and see if you still get the counter to increase.

Reduce it to a minimum so that you can capture a sniffer trace.

Internally we use Spirent equipment to run our tests.

Gilles.