Dashboard tile with number of hosts with CPU bigger than 80%?

[AntonioSousa]

In Data Explorer, if you want to have a list of hosts that have, eg. more than 80% of CPU for a certain timeframe, you put the following metric selector and select table:

builtin:host.cpu.usage:splitBy("dt.entity.host"):filter(series(avg,gt(80)))

Now imagine there are 10 hosts in the list and you want a single value in a dashboard with 10 in it. Been struggling to get this use-case done 🤣 Any ideas?
And yes, it's for a Managed environment...

[p_devulapalli]

Hi @AntonioSousa , how about using the table and pin it to dashboard ?

[rastislav_danis]

In fact table in data explorer is only way to really count records, but unfortunately, when pinned to dashboard that count is not displayed.

[rastislav_danis]

I'm afraid it's not possible, i was trying hard with similar task.

[Julius_Loman · Accepted solution]

@AntonioSousa what about this?

builtin:host.cpu.usage:splitBy("dt.entity.host"):filter(series(avg,gt(10))):count:splitby():last

[rastislav_danis]

It's not working - f.e. 5 hosts with cpu>0

[AntonioSousa]

@Julius_Loman 

I tried your selector and it was very volatile. It would sometimes work, but others not.

But I have worked with it, and came to this one, only without the last:

builtin:host.cpu.usage:splitBy("dt.entity.host"):filter(series(avg,gt(80))):count:splitby()

This one seems to work perfectly!

Thanks for the double splitBy() idea!

[PedroDeodato · Accepted solution]

If you skip the first splitBy() transformation, the expression can be simplified a bit and it still works.

The only requirement is that you apply the second splitBy() (in my case, the only one) before you apply the :count transformation, otherwise it won't work.

As a result, I got this:

builtin:host.cpu.usage:filter(series(avg,gt(80))):splitBy():count

Hope this helps!

[rastislav_danis]

I found out that you have to timeshift and then it's finally working:

builtin:host.cpu.usage:splitBy("dt.entity.host"):filter(series(avg,gt(0))):count:splitby():timeshift(-1m)

[rastislav_danis]

it's b/c data from all records are not collected in last time bucket, for other metrics collected with >1m interval you will need to perform bigger timeshift. Here is examle of what you can get if doing last():

[rastislav_danis · Accepted solution]

simple skipping :last() without timeshift() will also work for single value, but you need to let "Auto" or "Maximum" Fold transformation with global timeframe >1m:

[AntonioSousa]

Thanks all for your input! Several ways to do something that I believe is a good use-case 🤗
Which leaves me thinking that the efficiency of doing certain approaches in Data Explorer might lead to different usage of resources inside the Dynatrace platform.
Looks like @PedroDeodato metric selector is the most efficient, but does anyone have any idea which metric selectors are more efficient or not?

[rastislav_danis]

It was very useful for me as well, thanks for posting question! @PedroDeodato 's selector is ok, but you have to also check Fold transformation (for Single value visualization) to be in correct mode (Auto or Maximum).