ping input providing false/unreliable statistics with multiple url's

[redrumdk]

Hi Guys,

I have a big concern that the statistics collected by the native method of ping are incorrect.
The Maximum and Minimum response times are often the same for multiple different url's (ip's)
This makes me think that the averages might also be incorrect, as the min and max used would also reflex the average.

it looks like the statistics collected are somehow crossed and reused, and hence not correct.

i have attached a screenshot of the result collected, and color coded the entries i see where the number is the same for the differnt url's

https://ibb.co/0XStJwk

Debug logs provide no errors or additional information

would you please have a look at it.
thans for all the great work by the way

Relevant telegraf.conf:

[[inputs.ping]]
method="native"
urls = ["x.x.x.x",
"x.x.x.x",
"x.x.x.x",
"x.x.x.x", 
"x.x.x.x", 
"x.x.x.x", 
"x.x.x.x", 
"x.x.x.x", 
"x.x.x.x", 
"x.x.x.x",
"x.x.x.x"
] 
count = 10
timeout = 60.0

System info:

Telegraf 17.3, windows 10, windows 2016

Steps to reproduce:

1. Use the same config as pasted above, and check the database field maximum_response_ms and minimum_response_ms

Expected behavior:

Individual measurements

Actual behavior:

a pattern og the same measurements/statistics being reused for individual url's

[redrumdk]

@sjwang90
@ssoroka
:)

[sjwang90]

@redrumdk Thanks for bringing up the issue. Is this a recent behavior?

[redrumdk]

@redrumdk Thanks for bringing up the issue. Is this a recent behavior?

Yes, since latest release of telegraf 17.3. Ever since the Native method was Fixed.

[redrumdk]

same problem in v1.18.1

[hansenpansen]

I am experiencing the same with v1.18.1 on Oracle Linux 8.3. Raw capabilities have been enabled in the systemd unit-file for telegraf.

It does not matter if I have a single [[inputs.ping]] with multiple elements in urls, or multiple [[inputs.ping]] sections with a single element in urls.

[redrumdk]

I am also experiencing random packetloss reported for various ip's with telegraf, while I have been testing with blackbox exporter and prometheus, and ofcourse manually pinging in windows. And they report no packetloss.

Scenatio is:
i have 1 site im pinging which is permanantly down, and my theory is that some of the results from this is being mixed up with the rest of the result.
As i have been with the other result initially creating the case.
https://ibb.co/0XStJwk

[toomuchio]

Also experiencing this issue with Debian Linux, I've had to revert back to exec mode which does not have this issue.

It could be due to conflicting IDs generated for ICMP but that might be a lib issue
https://github.com/go-ping/ping/blob/master/ping.go#L100
https://github.com/go-ping/ping/blob/ff8be3320020b96856d3d48505bcf8926809752e/ping.go#L767

Those don't look completely RNG safe.

[redrumdk]

Also experiencing this issue with Debian Linux, I've had to revert back to exec mode which does not have this issue.

It could be due to conflicting IDs generated for ICMP but that might be a lib issue
https://github.com/go-ping/ping/blob/master/ping.go#L100
https://github.com/go-ping/ping/blob/ff8be3320020b96856d3d48505bcf8926809752e/ping.go#L767

Those don't look completely RNG safe.

Yeah, im on windows, going back to exec results in another error with that many urls im trying to use.

[toomuchio]

What error are you getting @redrumdk ?

For the devs this is a visualization of what native vs exec looks like.

[redrumdk]

What error are you getting @redrumdk ?

For the devs this is a visualization of what native vs exec looks like.

For the Native, you can check my screenshot and text in the first post. Im the one opening the case :)

For exec i have another case here.
#8704

Getting this error in the log, when i have too many IP's for Telegraf to handle
[inputs.ping] Error in plugin: , exit status 3221225794: X.X.X.X
And then it doesn't collect any measurement for the IP in x.x.x.x

[srebhan]

@redrumdk and @toomuchio does this issue still exist on recent versions of Telegraf?