scollector metric explaination

232 views
Skip to first unread message

Lin

unread,
Mar 3, 2015, 9:31:50 PM3/3/15
to open...@googlegroups.com
Hi, all,

I am considering recently about using scollector via here: 

Any experts can show me some metrics explaination guide for the collectors build-in with scollector?

Thanks,
Sun.

Kyle Brandt

unread,
Mar 14, 2015, 2:00:19 PM3/14/15
to open...@googlegroups.com
Sorry for the late reply, I had not looked at the list in a little bit. What information would you like?

We try to name metrics something like system.subsystem.metric . When you use it with Bosun bosun will collect some metadata about a lot of the metrics that scollector has. The main ones are:
 * The unit (i.e. bytes, bits, Page loads .. whatever)
 * The type (counter vs gauge)
 * A description telling you what the metric is about

All that metadata only comes into with scollector + Bosun right now used in conjunction with OpenTSBD. Currently, without Bosun, you can still get all the data but you won't get that metadata. In this situation you can look at the scollector source code to see what the metadata it is. We added metadata later in scollector's development, so not everything has metadata.

Scollector runs on Linux and Windows and has a good deal of metrics for both stacks.

Best,
Kyle

gary.w...@comparethemarket.com

unread,
Jun 28, 2016, 7:54:31 AM6/28/16
to OpenTSDB
Hi Kyle,
I have the same question as Sun. Where can I find a (plain english) description of the metrics which are outputted by scollector? Preferably with a good use case...
for Linux_loadavg_5_min for instance is there another metric I can divide it by to represent the number of cores or vCPU's of the instance, so I can get a true picture of the load?
i.e. what is machine_cpu_cores
what is linux_cpu_percpu
what is linux_disk_ios_in_progress

etc etc

Many Thanks

Kyle Brandt

unread,
Jul 8, 2016, 7:55:53 AM7/8/16
to OpenTSDB
There is no static catalog of this information. Generally we rely on Bosun to display this information since scollector sends metadata "desc" field for many of the metrics (but not all).

If you have a bosun instance and have scollector sending data to it (or it via tsdbrelay) the metadata is available, and you can see descriptions at http://bosun/api/metadata/metrics (orwhatever your bosun host is)

With just scollector you could so something hacky like:

$ /opt/scollector/scollector -p -d | grep desc | grep linux.cpu.percpu
2016/07/08 11:55:10 info: metadata.go:168: AddMeta for linux.cpu.percpu/cpu=0,host=ny-kbrandt02,type=user/desc: Normal processes executing in user mode.
2016/07/08 11:55:10 info: metadata.go:168: AddMeta for linux.cpu.percpu/cpu=0,host=ny-kbrandt02,type=nice/desc: Niced processes executing in user mode.
2016/07/08 11:55:10 info: metadata.go:168: AddMeta for linux.cpu.percpu/cpu=0,host=ny-kbrandt02,type=system/desc: Processes executing in kernel mode.
2016/07/08 11:55:10 info: metadata.go:168: AddMeta for linux.cpu.percpu/cpu=0,host=ny-kbrandt02,type=idle/desc: Twiddling thumbs.
2016/07/08 11:55:10 info: metadata.go:168: AddMeta for linux.cpu.percpu/cpu=0,host=ny-kbrandt02,type=iowait/desc: Waiting for I/O to complete.
2016/07/08 11:55:10 info: metadata.go:168: AddMeta for linux.cpu.percpu/cpu=0,host=ny-kbrandt02,type=irq/desc: Servicing interrupts.
2016/07/08 11:55:10 info: metadata.go:168: AddMeta for linux.cpu.percpu/cpu=0,host=ny-kbrandt02,type=softirq/desc: Servicing soft irqs.
2016/07/08 11:55:10 info: metadata.go:168: AddMeta for linux.cpu.percpu/cpu=0,host=ny-kbrandt02,type=steal/desc: Involuntary wait.
2016/07/08 11:55:10 info: metadata.go:168: AddMeta for linux.cpu.percpu/cpu=0,host=ny-kbrandt02,type=guest/desc: Running a guest vm.
2016/07/08 11:55:10 info: metadata.go:168: AddMeta for linux.cpu.percpu/cpu=0,host=ny-kbrandt02,type=guest_nice/desc: Running a niced guest vm.
2016/07/08 11:55:10 info: metadata.go:168: AddMeta for linux.cpu.percpu/cpu=1,host=ny-kbrandt02,type=user/desc: Normal processes executing in user mode.
2016/07/08 11:55:10 info: metadata.go:168: AddMeta for linux.cpu.percpu/cpu=1,host=ny-kbrandt02,type=nice/desc: Niced processes executing in user mode.
2016/07/08 11:55:10 info: metadata.go:168: AddMeta for linux.cpu.percpu/cpu=1,host=ny-kbrandt02,type=system/desc: Processes executing in kernel mode.
2016/07/08 11:55:10 info: metadata.go:168: AddMeta for linux.cpu.percpu/cpu=1,host=ny-kbrandt02,type=idle/desc: Twiddling thumbs.
2016/07/08 11:55:10 info: metadata.go:168: AddMeta for linux.cpu.percpu/cpu=1,host=ny-kbrandt02,type=iowait/desc: Waiting for I/O to complete.
2016/07/08 11:55:10 info: metadata.go:168: AddMeta for linux.cpu.percpu/cpu=1,host=ny-kbrandt02,type=irq/desc: Servicing interrupts.
2016/07/08 11:55:10 info: metadata.go:168: AddMeta for linux.cpu.percpu/cpu=1,host=ny-kbrandt02,type=softirq/desc: Servicing soft irqs.
2016/07/08 11:55:10 info: metadata.go:168: AddMeta for linux.cpu.percpu/cpu=1,host=ny-kbrandt02,type=steal/desc: Involuntary wait.
2016/07/08 11:55:10 info: metadata.go:168: AddMeta for linux.cpu.percpu/cpu=1,host=ny-kbrandt02,type=guest/desc: Running a guest vm.
2016/07/08 11:55:10 info: metadata.go:168: AddMeta for linux.cpu.percpu/cpu=1,host=ny-kbrandt02,type=guest_nice/desc: Running a niced guest vm.
Reply all
Reply to author
Forward
0 new messages