Question Advanced Monitoring - How do graphs work with multicore CPU?

[Will-NYESDigital]

‪The version of Obsidian I have is CentOS Linux 7.7.1908 (Core)

Hi there, I have a multicore system but would like to know does all CPU work as 300% or 100% for all 3 cores?

To do this I went to Advanced Monitoring then selected CPU

I am aiming to set thresholds once I have got the answer and that then leads on to my next question is that I have noticed the system idle is high 90% +. Can someone please advise me what this service is?

Thanks

 

[Ales]

System idle is not a service, it is the time your system is doing nothing. In other words, 90% idle means that 10% is actually in use while 90% is free.

Idle hands... well... in this case, the proverb about the devil doesn't apply, I think.

 

[Bitpalast]

‪
Hi there, I have a multicore system but would like to know does all CPU work as 300% or 100% for all 3 cores?

That's an interesting question. In the case of Grafana I think that 100% means "all CPUs counted together", because the idle limit is at 100%. Else, on a 3-core-system, it would be at 300%. On the OS level with "uptime" or "top", this is different. There a value like "3" on a 3-core-processor would mean 100%.

 

[seqoi]

That's an interesting question. In the case of Grafana I think that 100% means "all CPUs counted together", because the idle limit is at 100%. Else, on a 3-core-system, it would be at 300%. On the OS level with "uptime" or "top", this is different. There a value like "3" on a 3-core-processor would mean 100%.

I can confirm your findings. Example i have 4 cores vserver instance in one place (Hetzner). Plesk is showing 100% for all 4 cores total yes.

While in my Hetzner server panel interface (unrelated to Plesk) i see from 0 to 400%. Dynamically. Meaning it is showing treshold of 100% but if it goes up then maximum is up to 400% which makes sense (to me)

I guess this is just up to the infrastructure.

 

[Will-NYESDigital]

Excellent, thank you for the support with the combined core readings.

My next question is how do I remove the alert that keeps occurring due to the CPU idle time reaching 90%+ as I want to remove it from the graph completely to be honest. If I do remove that specific value once the site is refreshed I know that it will just come straight back.

Thanks again

 

[Bitpalast]

Are you sure you are getting an alert on idle time? I don't think that is possible, because the more idle time you have the better. Could you please post the "alert" here so that we can see what exactly the alert message reads?

 

[Will-NYESDigital]

Hi Peter, please read the following the error message that occurs regularly since updating to Obsidian

-
[Alerting] Total usage
CPU usage status is critical!
The current value is 90.7%.

Thanks again

 

[Bitpalast]

So we are not talking about idle time, but CPU usage. In that case you need to find out what your websites are doing that is consuming so much CPU power. Check your access_log and error_log files. Normally a high CPU load indicates brute force attacks and you may find many similar requests to the site, like constant Wordpress XML RPC requests. Are you using fail2ban to block such attacks?

 

[Will-NYESDigital]

Hi Peter the issue I am highlighting is the fact the 'cpu-percent-idle:value' data point is displaying 90%+ readings which is then setting of the alerts that the CPU usage is critical. Do I need to simply set my threshold for this 280% due to me having 3 cores?

Thank you in advance

 

[Ales]

Idle value should not be counted under "total usage". If it is, that would be a misconfiguration or a bug. It's more likely that the actual CPU usage is rising to 90%+, while the idle is below 10% at that point.

Also, in this case, I'd say that 100% being the maximum is correct.

Additionaly to what @Peter Debik suggested in the post above, are you perhaps using sysstat on your server? Look it up...

 

[Will-NYESDigital]

Hi @Peter Debik and @Ales I am using Fail2Ban but not using syssat on my server.

Please find attached the full error that is being displayed on one of my cores.

Thank you again for your support.

 

[Bitpalast]

It seems that "idle" is simply treated as all other curves. So the "Alert" message on idle should not be there. I think this is a minor bug with no impact on anything. You can safely ignore that "alert". It should rather say something like "Fantastic, almost nothing to do" and display a green checkmark or something. But the tool is not capable of that.

 

[Will-NYESDigital]

Hi all

What do you recommend setting my Thresholds for the following?

Apache & php-fpm CPU usage
nginx CPU usage
Mail server CPU usage
Plesk CPU usage

If you could also let me know some reasoning behind it to that would be great.

The term: Exact value (µs) is new compared to the Health Monitoring service so this has thrown me.

Thanks for your time as always.

 

[IgorG]

What do you recommend setting my Thresholds for the following?

Apache & php-fpm CPU usage
nginx CPU usage
Mail server CPU usage
Plesk CPU usage

Could you please clarify why default settings of thresholds are not suitable for you?