High CPU on Cacti

[G1]

I am using the ONTAP SDK scripts to capture storage data. As soon as I configure 10 or more data sources, the CPU of Cacti host goes upto 100% and the processes consuming CPU are the Cacti perl scripts.

Is there any way or any customized configuration which could be done, to reduce CPU consumption ?

Appreciate any help on this.

[gandalf]

Are those ONTAP scripts CPU intensive? Or is your perl wrapper doing something insane? Cacti itself does not usually take many CPU cycles for running a script.
R.

[G1]

Hello gandalf,
Thanks for the reply.

What i have noticed is the OnTap scripts for VOlume Latency are taking up the CPU. As soon as I disable these scripts, and enable only the Volume IO scripts of OnTap, the CPU consumption is low.

For now, I have disabled the Volume Latency scripts. Pls let me knowif you have any idea.

THanks,
G1

[G1]

I have to correct my earlier message. The CPU utilization goes high even for the Volume IO volumes. If more than 8 or 10 volumes are configured the CPU reaches 100%.

Can anyone pls help.


Regards,

[G1]

Hello Gandalf,

Could you pls let me know more about the perl wrapper. How to identify whether this is an issue with the ontap sdk script or the perl wrapper?

[gandalf]

Run "top" or the like.
RUn the "native command" and observe top.
Run the perl script and observe top.
If results differ to a great extend, it's the perl script.
R.

[G1]

Thank you for the suggestion. I will get check and back to you soon.

[G1]

Tried executing the perl scripts manually. Still it takes up the CPU.
This would mean that there is some other issue due to which the CPU is consumed upto 100%. Is there any fine tuning parameter which could be tweaked to limit the cpu usage ?

[gandalf]

As we don't know that script and your environment, it's quite impossible to help
R.

[rujipars]

If that kind of information is necessary, would you try putting the routine measuring job to another machine and place the output (may be a simple text file) on cacti server then use some simple method to read input from that file?

I've applied this kind of asynchronous operation for some measurements that take a considerable time like a series of pings and observe the packet loss to several targets worldwide. I did that with MRTG but we can apply the same concept with cacti.

[TheWitness]

The Ontap scripts are CPU intensive. I prefer using snmp to monitor the NetApps. NetApp has a fairly comprehensive agent. However, the quality of the Agent depends on it's age.

TheWitness

[G1]

hi TheWitness / All , thanks for your response.

Are you able to monitor NetApp at the volume level instead of LUN level?

Could you let me know what exactly and how do you use ( any link to the template) ?

[mkalle]

i still use ONTAP for fetching some things, but not all data is avaible through snmp. but Volume stats are.

i reduced the poller runtime from 50sec to 15 sec

[G1]

hi TheWitness/All,

It is observed that as soon as more than 10 or 15 data sources are configured, the resource utilization shoots up. CPU reaches 100%. If I leave it as it is then after a day or so, the system is hung giving out of memory message on console.

There may be a memory leak in the script or some more RAM is needed.

Any inputs appreciated.

Using ONTap SDK script which internally uses the volume.xml script.

Pls help.

REgards,

[G1]

Additional inputs: The system goes in the hung state and is needed to hard reboot.

ON hte console it says free swap = 0kb and gives OOM Out of memory messages and kills (randamly) any processes one by one.