Forum Discussion

Caio_178191's avatar
Caio_178191
Icon for Nimbostratus rankNimbostratus
Nov 27, 2015

CPU showing controled peaks (5 in 5 minutes or 10 in 10 minutes)

Hi.

 

I had experience with some F5s with differents BIG-IP versions(up to 11.5.x) and hardware.

 

Some BIG-IPs show CPU monitor giving a peak to 100% 5 in 5 minutes for some seconds.

 

I also had experience with another versions and hardware show peaks 10 in 10 minutes and I also see some BIG-IPs without this peaks.

 

For example, right now I'm in a customer implementing 2x4200 in cluster with v11.5.1 HF10. As I said both hardwares are in a Sync and failover group with same big ip version and load balance objects. The only difference between is the self ips. The systems are working in active/active mode. Each one have almost the same number of connections and SSL processes as the other.

 

Right now the CPU monitor from one of the systems is showing a peak in each 5 minutes, and the other one is not showing this behavior.

 

Please, any explanation for this controlled peak? It look like there is nor rules for that, or I'm missing something.

 

Thanks.

 

2 Replies

  • seen this mentioned before and thought it was some bug, but can't find a link or something. you could try with the linux top command to find the cause. other thing might be to involve support.
  • The 5 minute interval makes me think of this:

     

    https://support.f5.com/kb/en-us/solutions/public/15000/600/sol15606.html

     

    Also I recommend following the link at the bottom of that page to SOL15003 regarding HT-Split. When looking at CPU spikes in 11.5.0 and greater its important to keep in mind that data plane and control plane tasks have been split up to even and odd numbered logical cores, so its important to look at which cores are spiking. Hope this helps.