inputs.ping - When one target url not responding it stops and dont ping the rest

[kawaiipantsu]

Relevant telegraf.conf:

[[inputs.ping]]
   urls = [
       "target1",
       "target2",
       "target3",
       "target4",
       "target5",
       "target6",
       "target7",
   ]
   method = "exec"
   count = 2

System info:

# uname -a
Linux telegraf 4.14.138-rancher #1 SMP Sat Aug 10 11:25:46 UTC 2019 x86_64 GNU/Linux

# telegraf --version
Telegraf 1.12.5 (git: HEAD f09d4023)

Steps to reproduce:

Have a long list of targets (url) target 1 to 7.
Now pull target3 down and NS record, and telegraf ping will only update target1 and 2 in influxdb.
The rest will shown as down.

I have tested both with "method" set to native and exec.

Expected behavior:

Target 1 to 2 - up and metrics
Target 3 - down
Target 4 to 7 - up and metrics

Actual behavior:

Target 1 to 2 - up and metrics
Target 3 - down
Target 4 to 7 - down

Additional info:

Also it's irritating that when a host is not responding that we don't set package loss to 100%.
Just a side note :)

[danielnelson]

Fixing this for 1.12.7.

Unrelated to your bug, could you try switching to method = "native"? Our long term plan is to remove the exec method, once native has proven itself. On Linux you will need to modify some permissions to make it work.

Also it's irritating that when a host is not responding that we don't set package loss to 100%.

I'm assuming you mean when there is a dns lookup error? This has been requested before but since we aren't sending any ping packets when this error happens I didn't want to set this value. Maybe it would make sense to have a processor that can set a default value if a field is unset.

[kawaiipantsu]

I tested your commit and yes it works as it should. It now continues on dns error. Also nice you changed the tag, somehow "url" just seemed misleading.

[kawaiipantsu]

Fixing this for 1.12.7.

Unrelated to your bug, could you try switching to method = "native"? Our long term plan is to remove the exec method, once native has proven itself. On Linux you will need to modify some permissions to make it work.

Also it's irritating that when a host is not responding that we don't set package loss to 100%.

I'm assuming you mean when there is a dns lookup error? This has been requested before but since we aren't sending any ping packets when this error happens I didn't want to set this value. Maybe it would make sense to have a processor that can set a default value if a field is unset.

Hey, i have not had time to test it with native sadly, also we had to get it working on that day I filed the issue so I ended up coding a daemon in php running fling and doing the same. Perhaps we shift back to telegraf but for now this is working perfectly also expanded it to look at url's and http return codes.

Yeah the default value would properly be a good idea, like "what to return if all fails" but for what metric. It could also just be that if all fails you at least set package loss to 100%. This is what I have done in our own daemon as we see/use that indicator for way to express if a host is down or unreachable to..

[danielnelson]

Also nice you changed the tag, somehow "url" just seemed misleading.

Just to avoid confusion, I didn't do this. I agree that url is wrong but I left it for now to keep backwards compatibility.