Float item type

[simsaull]

Hello all!
we use a custom script to monitor our NAS that outputs data to zabbix_sender.
The item memory_free was set as a Numeric, with "B" unit.
After upgrading the filer to a bigger model with more RAM, the script outputs this value for memoy_free : 7.15443e+09
So Zabbix rejects that value as it's not numeric. I tried to change the type to float and it doesnt work either.

When I try to send that value manually with zabbix_sender in verbose mode all I get is :
zabbix_sender [13327]: DEBUG: answer [{
"response":"success",
"info":"Processed 0 Failed 1 Total 1 Seconds spent 0.000055"}]
Info from server: "Processed 0 Failed 1 Total 1 Seconds spent 0.000055"
sent: 1; skipped: 0; total: 1

Any idea why?

Thanks.

[steveboyson]

Why don't you simply change the script and let it output kB or MB? Stay on "B" in zabbix but set a custom multiplier in Zabbix to re.get the correct values
If you receive the value in kilobyte, use 1024

[simsaull]

Hello steveboyson, thanks for you answer.

I do not want to change the script simply because it is used to monitor different NAS systems, small ones and bigger ones. Using a float type for "memory_free", Zabbix still accepts integer values returned by the script when monitoring the small filers, so it seems like a good solution.

So back to my question : what is wrong about "7.15443e+09" being a float? This would be a valid value for float variable in C right?

Thanks again.

[EnigmA-X]

What version of Zabbix are you using on client and server?

You wrote that you are using Numeric for the item type. Is that Numeric (unsigned) or Numeric (float)?

[simsaull]

Hum, yes I realized that I have not been very accurate.

I was using "Numeric (unsigned)", worked perfectly, now I'm trying to use "Numeric (float)".

I am running Zabbix 1.8.6 on the server and v1.8.6 (revision 20932) on the client.

Thanks

[EnigmA-X]

Ehm, I'm not a C-programming expert, but...

A float is 32 bits long, has a precision of 7 digits. While it may store values with very large or very small range (+/- 3.4 * 10^38 or * 10^-38), it has only 7 significant digits.

For the large numbers you are monitoring, I really urge to start using integers instead of floats. It makes no sense to me, to use the precision of floats for such large numbers, because imho the 7 significant digits is a huge drawback in this specific case and integers will give you much more precision.