The problem of Division by zero

[xiaoquwl]

Hi all,

I have a problem to ask.
One Item I used in zabbix is calculated, it ‘s formula is last("To_SAA-CSJCZ1_UDP-RTTSum") / last("To_SAA-CSJCZ1_UDP-NumOfRTT")
But in the log file “zabbix_server.log”
there is a error like
“error: Division by zero. Cannot evaluate expression [0/0]”
How can I avoid this error by adding some judgment, I mean if the last("To_SAA-CSJCZ1_UDP-NumOfRTT") is zero, I just set this formula’s result is “no data”, or if it’s not, then it’s will be calculated. can I ? How can I do that ?
Is there anybody can help me ?

That will be very thankful for you information.

[marcel]

adjust the formula so you never divide by zero:

last("To_SAA-CSJCZ1_UDP-RTTSum") /( last("To_SAA-CSJCZ1_UDP-NumOfRTT")+0.1)

[toks]

Hi, is there a better answer for this problem? To sum '0.1' doesn't seem an alternative to me. Is there a function that I can use in calculated items that returns zero if there is a division by zero?

Thanks in advance.

[nelsonab]

Use the Lua patch. :-)

[rogeriomgatto]

This formula below works for me, it changes 0 to 1 but doesn't change anything if divisor is not 0.

last("X") /( last("Y") + count("Y",#1,0) )

[Pada]

This formula below works for me, it changes 0 to 1 but doesn't change anything if divisor is not 0.

last("X") /( last("Y") + count("Y",#1,0) )

Thank you! That is a fantastic way of doing it!

Unfortunately that method does not work when you devide by the sum of 2 items.
eg. last("X") / ( last("Y") + last("Z") )
So for that case I'm doing:
last("X") / ( last("Y") + last("Z") + 0.000001 )

[clarkritchie]

We also get a lot of "Division by zero. Cannot evaluate expression" errors based on a formula that's on an item prototype in a discovery rule. Here's what a coworker came up with:


100*last(If.32.ifInOctets.["{#SNMPINDEX}"],0)/last(If.32.ifSpeed.["{#SNMPINDEX}"],0)


If.32.ifSpeed.["{#SNMPINDEX}"] is sometimes 0, I think maybe when the interface is disabled?

Anyhow, I believe the ",0" is not needed at all as we are not time shifting the results of the last() function, thus it does nothing.

Based on reply #5 (above), the formula would become:


100*last(If.32.ifOutOctets.["{#SNMPINDEX}"])/(last(If.32.ifSpeed.["{#SNMPINDEX}"]) + count(If.32.ifSpeed.["{#SNMPINDEX}"],#1,0))


But I still get div by 0 errors. Advice?

[Zabbix_User_1138]

We also get a lot of "Division by zero. Cannot evaluate expression" errors based on a formula that's on an item prototype in a discovery rule. Here's what a coworker came up with:


100*last(If.32.ifInOctets.["{#SNMPINDEX}"],0)/last(If.32.ifSpeed.["{#SNMPINDEX}"],0)


If.32.ifSpeed.["{#SNMPINDEX}"] is sometimes 0, I think maybe when the interface is disabled?

Anyhow, I believe the ",0" is not needed at all as we are not time shifting the results of the last() function, thus it does nothing.

Based on reply #5 (above), the formula would become:


100*last(If.32.ifOutOctets.["{#SNMPINDEX}"])/(last(If.32.ifSpeed.["{#SNMPINDEX}"]) + count(If.32.ifSpeed.["{#SNMPINDEX}"],#1,0))


But I still get div by 0 errors. Advice?

Works if you add double quotes around the counted itemkey, like this:
100*last(If.32.ifOutOctets.["{#SNMPINDEX}"])/(last(If.32.ifSpeed.["{#SNMPINDEX}"]) + count("If.32.ifSpeed.["{#SNMPINDEX}"]",#1,0))

I haven't tested it, but you might need to escape the double quotes within the quoted itemkey, with backslashes like this:
100*last(If.32.ifOutOctets.["{#SNMPINDEX}"])/(last(If.32.ifSpeed.["{#SNMPINDEX}"]) + count("If.32.ifSpeed.[\"{#SNMPINDEX}\"]",#1,0))

I know this is a very late reply, but others with the same issue might benefit.

[wally.hall]

I know this is an old thread, but it helped me today and I successfully "overcame" the issue of division by zero for percentage calculated items by further abusing the limitations of floating point arithmetic...

For the example typical percentage equation:

READING / MAXIMUM * 100

...I've calculated everything several orders of magnitude larger before shrinking the output back down:

(READING * 10000000) / (MAXIMUM * 10000000) / 100000

(Note the final division is just me optimizing out the * 100...)

This seems to round out to zero in Zabbix - or at worse 0.001.