Forum Discussion

DevBabu_124276's avatar
DevBabu_124276
Icon for Nimbostratus rankNimbostratus
May 05, 2015

icontrol bigsuds

In my python script I am checking if the nodes are present or not and if not create the node. Similarly, I am checking if the pool is present or not and if not present create the pool.

 

There are nearly around 1000+ nodes and 400+ pools.

 

Do the large loop for checking nodes and pools will highly impact CPU usage and memory in LTM. Do creating 50+ pools and 100+ nodes from the script will have any high impact on CPU and memory in LTM.

 

I am testing in my LAB box where I don't have much configuration as in prod boxes, so not sure how much will it impact in prod.

 

Any input will be highly helpful.

 

Thanks.

 

1 Reply

  • Joe_Pruitt's avatar
    Joe_Pruitt
    May 05, 2015

    The batch methods were designed to create multiple objects, but there are some theoretical limits which mainly are centered around how large the configuration is for each object (meaning how large the SOAP payload is that's sent and parsed). Querying statistics is somewhat of a costly operation as the SOAP/XML response is quite large while querying the lb_method of a virtual server is much lighter. It also depends on how hot the BIG-IP is running (meaning how much of the CPU is being used). If you are at or near 100% capacity on your device, then I'd say splitting things up into smaller chunks is a good idea. It's mainly the processing/parsing of the request/response that's the issue.

     

    As a general rule of thumb, I'd say to batch things in 100 increments is not a bad way to go. If you have 110, then include them all in one, if you have 250, then split them into a couple of calls.

     

    But, ultimately it will depend on how much your device is being used. If you are running light, then you can increase the numbers in each batch.

     

    Hope this helps...

     

    -Joe