Netdata warning thresholds

[danb35]

Wondering what thought process went into setting the warning thresholds for netdata, as it seems like I'm getting lots of warnings in my email for things that shouldn't be a concern at all. For example, RAM utilization > 90%. Is that a problem? Is it not rather to be expected when ZFS caches so aggressively? And disk utilization > 90%? Again, it seems like that would be expected when a scrub is running. I had a couple dozen netdata alerts in my admin email this afternoon, and they were all of this sort. Why does the system think I should care?

 

[Apollo]

When Netdata was introduced with 11.2 Beta, I was getting such warning on a daily basis.
As you said, scrubbing would trigger alerts. Using Nextcloud and having files transfered would trigger LAN alerts.
Now that 11.2 Release is out, I almost no longer get those alerts anymore.
As a matter of fact, I haven't received any alarts since Dec 18, which seems about the time I have upgraded to Release.
Before I used to have packet storms around 6000% or more in 10s.

At first this was a scary thing because of how it was reported, either as a warning or critical, as if the end of the world.
Now, everything is silent.

 

[Redcoat]

I just figured it was AI learning my level of sensitivity...

 

[SweetAndLow]

I filed a bug for the RAM usage warning. It got attention and will be fixed. They also mentioned that in later releases there will be a UI to set the thresholds to whatever you want. Another funny alert I got was when I upgrade from 11.2 to 11.2-U2 I got a warning about disk usage because it was writing so much durning the upgrade to my boot device.

 

[SweetAndLow]

Woke up to 48 emails about disk busy warnings from a scrub. I'm going to disable netdata now.

 

[garm]

Does anyone know what would not constitut an unscheduled reboot? Those emails bugs me too.. I get them for updates and manual UI reboots.

 

[garm]

Also, I did have lots more emails in the beginning of RC2. Now with Release I get way fewer. Is there learning code in Netdata?

 

[danb35]

Think you've got my threads confused. But in my case, "unscheduled" seems to mean "uncommanded"--that is, the system reboots on its own. I have not received those emails when I told it to reboot.