iRules and ASM

Hello,

 

 

You can actually use an HTTP class to filter which requests are sent to ASM and which are sent directly to a pool.

 

 

This is described in the 9.2.3 ASM release notes: (Click here)

 

 

The general idea is that you should be able to perform most decisions for which requests to send to ASM and which are sent to the pool using the HTTP class filters.

 

 

There have been and will be some pretty cool changes to the functionality for using iRules with ASM though. I'm planning on writing a solution to give more detail on this once things settle down.

 

 

If you have any questions on using rules with ASM post again here. If you run into any significant issues, try contacting the ASM support team using tssupport@f5.com. Please include a complete problem description, a serial number for the unit and a tech.out so we can review your configuration.

 

 

Thanks,

 

Aaron

Great,

 

 

Thats working out fine.

 

 

Also interested in measuring performance with ASM, the GUI really leaves a lot to be desired so I am digging in elsewhere.

 

 

I read about discrepancies between the UNIX tools (top, etc) and what is actually going on.

 

 

I would like to try to measure some basics - with ASM and without, specific costs to transactions with ASM etc. What are people doing - traps? iRules?

 

 

 

Hi,

 

 

The simplest method might be to use a benchmarking program to perform load testing against your VIP with and without the HTTP class associated with the VIP (without the class, requests are not sent to ASM). Apache bench (ab) is a pretty simple benchmarking tool installed on BIG-IP. Note that you can't test from the BIG-IP to a VIP on the same unit though (this will be possible in 9.4, but not in 9.2.x).

 

 

If you want to measure the resources used by an iRule (whether it's the internal _ASM_clientside, _ASM_serverside or custom rules) you can use the timing feature. For examples try searching the 9.x rules forum for timing. AskF5 also has a couple of related solutions: SOL6228 ().

 

 

SOL3242 (Click here) and SOL6583 (Click here) detail CPU and memory allocation for TMM in 9.x and explain why top and ps don't show valid data for TMM resource utilization.

 

 

The performance graphs actually do show accurate statistics (globally) for TMM memory and CPU utilization. It might be possible to use rrdtool to gain access to the raw data used to generate the stats. I don't have any ready examples, but you can look at the rrdtool man page or contact support for more info.

 

 

Hope this helps...

 

Aaron