Packet loss data in performance graph from check_icmp

[perric]

Hi,

We have our hosts set up with a slightly modified check-host-alive (which warns at 300 ms / 20% packet loss and goes critical at 500 ms / 50% packet loss) with check_interval = 5, retry_interval = 1, max_check_attempts = 2.

Code: Select all

$USER1$/check_icmp -H $HOSTADDRESS$ -w 300.0,20% -c 500.0,50% -p 5

When I look at performance data for this host, I get some interesting values for packet loss such as in the attached = 17.87% packet loss for 2 minutes, then 3% packet loss for 2 minutes, then back to no packet loss. However, if we are only doing 5 pings per the check, shouldn't the packet loss be either 0%, 20%, 40%, 60%, 80%, or 100%? How is Nagios calculating this number?

I am asking this as we see value in capturing packet loss at a more granular level (ideally as these are data center hosts we'd like to alert on ANY packet loss, but don't want to ping the hosts hundreds of times on each check to catch small levels of packet loss)

Thanks

[benjaminsmith]

Hi @perric,

However, if we are only doing 5 pings per the check, shouldn't the packet loss be either 0%, 20%, 40%, 60%, 80%, or 100%? How is Nagios calculating this number?

Good question. That is correct ,however, Nagios XI uses a Round Robin Database to store performance data. This type of database is used to store large amounts of time series data in an efficient manner, and utilizes sampling techniques to reduce the amount of storage space required. So the data points on the graph take into account the sampling, averaging, interpolation of data by RRDtool and are not exactly what was returned by the plugin.

Let me know if you have any other questions.

For More information about RRDtool:
https://oss.oetiker.ch/rrdtool/index.en.html

[perric]

OK, thanks for the explanation.

[benjaminsmith]

Hi @perric,

OK, thanks for the explanation.

No problem. Did you have any other questions or shall I close this thread?

Thanks.

[perric]

Good to close, thanks