How Cacti Polling behaves for Down Devices ?

[abhisekp34]

How Cacti behaves for Down Devices ?

In Our Cacti there are around 250 + devices were in Down State. (not Disabled)

We saw issue in Poller performance for Live Devices. So, Some graphs were stopped plotting.

After Removing 100 Down Devices from Cacti, We saw improvement in Cacti Polling Performance. And Graphs were started plotting required data.

So, Does Down Device have Impact on Polling Performance in Cacti ? If so, How ?

Dose Spine Poller keep a separate Thread & Process for Down Devices ?

[Osiris]

What data collector?

[abhisekp34]

one Data collector, as it has only One Poller configured. (main poller)

[Osiris]

cmd.php or spine?

[abhisekp34]

It is Spine.
in Poller:- 8 Processes & 50 threads

[TheWitness]

I've mentioned the issue with processes and threads. If you want to see why this is bad, you should write a MySQL top command to see how many cycles are wasted inside the database "waiting" on other threads.

Here is a raw version of it.

Code: Select all

#!/bin/sh
while true; do
  clear
  mysql -e 'show full processlist' | egrep -v "(system user|Sleep|show full processlist)"
  sleep 1
done