Fix metric creation when node is offline in jenkins input

[danielnelson]

closes #6578

Required for all PRs:

• Signed CLA.
• Associated README.md updated.
• Has appropriate unit tests.

[danielnelson]

@ctNb Would you be able to test out this fix with your offline node? My Jenkins setup only has a single node right now. Here are some builds, let me know if you need a different one:

• telegraf-1.13.0~c2ee8640_linux_amd64.tar.gz
• telegraf_1.13.0~c2ee8640-0_amd64.deb
• telegraf-1.13.0~c2ee8640-0.x86_64.rpm

[ctNb]

i just tested the telegraf client on ubuntu and it seems not to work as expected. The offline node still doesn't show up.
I don't know how the accumulator exactly works, but is it possible that a metric is not getting listed when no fields are added? Because "status" is a tag and all other fields are in my case null (the API snippet in the issue shows this).
I am still testing some cases to be sure there is no error on my side

[danielnelson]

This is true, metrics without any fields are completely removed as they pass through Telegraf. I was expecting that the response_time field would still remain though. Let me know if you figure anything out and in the meantime I'll work on getting a better Jenkins test environment together.

[ctNb]

i am also setting up a clean test environment. I've installed the latest jenkins docker image and just added a node, which is not even existing. Jenkins shows this node in the GUI and API also as offline. When running now telegraf, the node is also not shown.
Maybe the solution is to add something like in the "http_response" plugin, which contains a "result" tag and a "result_code" as a field.

[danielnelson]

Here is my output from an offline slave, so it seems that the response time monitor doesn't stick around for long, or maybe it stays only after some failures.

{
	"_class": "hudson.slaves.SlaveComputer",
	"absoluteRemotePath": null,
	"actions": [
		{}
	],
	"assignedLabels": [
		{
			"name": "slave"
		}
	],
	"description": "",
	"displayName": "slave",
	"executors": [
		{}
	],
	"icon": "computer-x.png",
	"iconClassName": "icon-computer-x",
	"idle": true,
	"jnlpAgent": false,
	"launchSupported": true,
	"loadStatistics": {
		"_class": "hudson.model.Label$1"
	},
	"manualLaunchAllowed": true,
	"monitorData": {
		"hudson.node_monitors.ArchitectureMonitor": null,
		"hudson.node_monitors.ClockMonitor": null,
		"hudson.node_monitors.DiskSpaceMonitor": null,
		"hudson.node_monitors.ResponseTimeMonitor": null,
		"hudson.node_monitors.SwapSpaceMonitor": null,
		"hudson.node_monitors.TemporarySpaceMonitor": null
	},
	"numExecutors": 1,
	"offline": true,
	"offlineCause": {
		"_class": "hudson.slaves.OfflineCause$LaunchFailed"
	},
	"offlineCauseReason": "This agent is offline because Jenkins failed to launch the agent process on it.",
	"oneOffExecutors": [],
	"temporarilyOffline": true
}

Adding another field will do the trick. I think for now the safest change, just to address the issue, is to add the numExecutors field. We could think about other fields in the context of a feature request.

[danielnelson]

Example output:

> jenkins_node,arch=Linux\ (amd64),disk_path=/var/lib/jenkins,node_name=master,status=online,temp_path=/tmp disk_available=517427200,memory_available=1216356352,memory_total=2101669888,num_executors=2i,response_time=0i,swap_available=0,swap_total=0,temp_available=517427200 1573546293000000000
> jenkins_node,node_name=slave,status=offline num_executors=1i 1573546293000000000

Updated builds:

• telegraf-1.13.0~303518b2-0.x86_64.rpm
• telegraf-1.13.0~303518b2_linux_amd64.tar.gz
• telegraf_1.13.0~303518b2-0_amd64.deb