CPU to 100% in some cores ! CCR1036-12G-4S

Hello fellas!

Did you know why my cloud core don’t balance the charge over all the other 35 CPUs?.

Running the last ROS 6.13 CCR1036-12G-4S.

Thanks for your help in advance!.
Screen Shot 2014-05-23 at 1.20.36 PM.png

Some things cannot be balanced over multiple cores.
Among things that can only run on 1 core: the routing protocol processes (routing itself is multithreaded) and scripts.
Tool → Profile can give you a hint on what’s causing load on the CPU.

Some of these singlethreaded things will ultimately be made multithreaded in a later RouterOS version. But some things just cannot.
Imagine the task of putting one screw in a piece of wood. Having 36 hands and 36 screwdrivers will not make this task go any faster than a single hand and a single screwdriver.

Im concern because in my monitoring system my consume graphs shows like the equipment its processing to much and thats why its loosing connectivity and don’t showing any more the traffic passing . Thats happened me before in some UBNT devices, but also happened long time ago with CACTI when we are using a 32 bit for graphing and we was passing more than 32 bits pointer could process… but know we are using zabbix for monitoring and I think this guys program everything in 64 bits…


CCR1036-12G-4S



:confused:
Screen Shot 2014-05-30 at 11.36.25 AM.png

Well, what does the Routerboard itself say about CPU usage?

As for any SNMP monitoring of bandwith: Always select the HC / 64bit counters.
Cacti has absolutely no problem with 64bit SNMP counters.

Hello ppwicho,

perhaps we should talking about some tales and curious reports
made by some users here inside the forum

Many “Core” CPU:
You see it from the false side! It is not like a normal Intel CPU
with real CPU cores and the same number once again with HT
Technology! So you will see CPU count on the left side in your post
and not a Core counter! The Tilera CPU´s are many real and tinny CPU´s
that are meshed together and if one of this cpu´s is doing or working on an
process it could not be given to the next CPU.

The “idle” Process:
Otherwise very often some users report here in the forum that a process called
“idle” is taking much more then 80 - 90 % of the CPU usage, idle means this CPU
is doing nothing!

Real code writing and exceptions or user wishes:
The Tilera CPU´s from the Tile GX series are coming sorted
with many real and tiny CPU´s and not CPU cores, let us imagine
and talk about the TileGx36, coming with 36 of them.
So now you are programming the entire system with 16
for the WAN and 20 for the rest of the entire OS and one
CPU (“Core”) is now processing one service of the OS and the
rest should be doing nothing, and so it comes, that you will see
this “behaviours” like you are now reporting to us.

Not all services, features or options or functions will be able to
spread over many CPU´s or Cores like the most user want to see
or imagine. Let us thing this will be do able, what you personally
think the most peoples want to do? They want to spend all “Cores”
to their favourite RouterOS Service and than we will see what kind of
lame duck this OS or the entire CCR will be and this is not the real thing
of a RouterBoard I think really, because this would be bringing more trouble
and discussions about the device that must not be done.

So I think really in some month the RouterOS for Tile Gx platform will be finished up
and then it will be running smooth and fast for the most of us.

Thanks for your answer ,

I just change my OIDs from IF-MIB::ifOutOctets to IF-MIB::ifHCOutOctets. Also I disable the global queues trees and using back Simple Queues until RouterOS 7 be release.

Thanks!.

rOS 7.15.3 nad we are in square one…
Snímek obrazovky 2024-08-21 224354.png