Update go-ping to latest version #8771
Conversation
![telegraf-tiger[bot]](https://avatars.githubusercontent.com/in/94907?s=60&v=4){kind=small}
|
@redrumdk If you would like to try out this change, here are the Telegraf exectuables: https://app.circleci.com/pipelines/github/influxdata/telegraf/3092/workflows/755bf8fc-366a-4d4a-9228-c692611efb8a/jobs/75778/artifacts If you do get a chance to try it, let us know. We would love to see if the pending changes to go-ping helps the problem you are facing in #8704. Thanks! |
Yeah, just tried the pull request version. With config [[inputs.ping]] List of urls to pingurls = [ |
sspaink
changed the title
|
@redrumdk thank you for trying out the change, over the weekend @ssoroka made progress on the upstream change and got it successfully working and merged. If you are able to can you try the newest package? https://app.circleci.com/pipelines/github/influxdata/telegraf/3119/workflows/6b0a7e09-29f4-467c-9dcd-0a634fe791e3/jobs/75947/artifacts |
just tried it, EXEC still gives me errors with 114 test ip's and count set to 4 Switching to Native, now my permanently down IP's are showing 100 percent_packet_loss, which is great. |
|
is it possible they've actually got 50% packet loss? A side effect of this update is that we match the exact packets. eg: |
It's a constant 50% packetloss, no it's not possible :) if you look at my screenshot, i have crossed out the IP's, but they are all unique IP's all showing 50% packets loss, with 2 transmitted packets, and 1 received, and constantly as long as telegraf with native is running |
|
@redrumdk are you specifying a Timeout on your config for the plugin while running in native mode? if so, what is it? |
|
I am not :)
…On Wed, 10 Feb 2021, 15:52 David Bennett, ***@***.***> wrote:
@redrumdk <https://github.com/redrumdk> are you specifying a Timeout on
your config for the plugin while running in native mode? if so, what is it?
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#8771 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AMYJE5IQICKDJEBBL2XOLDTS6KMSZANCNFSM4WX3VTDQ>
.
|
|
@redrumdk so I was able to replicate this behavior on windows. I used 100 URLs and had random 50% packet loss for a decent chunk of them. however, after I bumped up the timeout the issues went away. I'd suggest configuring a high timeout and seeing if this helps. For me a 10 second timeout resulted in 0 packet loss. This is for native mode and the latest telegraf (including this pull request) I believe what's happening here is that the timeout default of 1 second just isn't enough time to do 4 sends/receives for whatever reason, hopefully. |
|
Nice, thank you. I will try as soon as i can.
How about the Count parameter, does that work for you on windows using
native ?
…On Wed, 10 Feb 2021, 18:39 David Bennett, ***@***.***> wrote:
@redrumdk <https://github.com/redrumdk> so I was able to replicate this
behavior on windows. I used 100 URLs and had random 50% packet loss for a
decent chunk of them. however, after I bumped up the timeout the issues
went away. I'd suggest configuring a high timeout and seeing if this helps.
For me a 10 second timeout resulted in 0 packet loss. (might be able to go
lower, not sure)
I believe what's happening here is that the timeout default of 1 second
just isn't enough time to do 4 sends/receives for whatever reason,
hopefully.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#8771 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AMYJE5PNSO6IHHL3FJBHJE3S6LAGXANCNFSM4WX3VTDQ>
.
|
|
@redrumdk yes. I was using a count of 4, so 4 packets were sent/received before reporting metrics. Note that the timeout needs to cover the amount of time it will take to send and receive all 4 of them. |
Damn, seems like your timeout settings worked! Tried it for EXEC method as well, but that didn't help. |
|
@redrumdk curiously I was able to run it with exec with no issues, so that still might be something to look into. But glad to hear Native is working! and that's correct, I'm thinking go-ping might want to change so that the timeout is only for one packet send/receive as opposed to all of them. |
it worked for you with 100 ip's and count of 4+ without getting errors in the debug log for exec ? or errors = 100 in the database ? |
|
i tested today using the latest official download, and the nightly build (1.18) and they still have problems. |
|
@redrumdk you're referring to the exec method correct? and yes. I had 100 urls / count of 4 and didn't have issues, but perhaps I need to increase that. I saw your issue about running in exec mode, haven't gotten a chance to dive into it yet. |
(cherry picked from commit f2cf447)
DON'T MERGE, for testing only to create temporary artifacts to give community a chance to test changes.Fix #8704
A pull request by @ssoroka has been merged go-ping that should resolve concurrency issues with inputs.ping plugin.