Implement basic metric with prometheus by t0kieu · Pull Request #1178 · fireice-uk/xmr-stak

t0kieu · 2018-03-21T02:29:39Z

Hi all,
This PR implement basic measuring with Prometheus metrics technology.
If you enable metrics while building app, then metrics are available on host:port/promethes

counter - total submitted result based on the type.
gauge - difficulty
gauge - hash rare total
gauge - hash rate per thread with device type.

Example of output:

# TYPE xmrstak_submitted_results_total counter
xmrstak_submitted_results_total{type="rejected"} 0.000000
xmrstak_submitted_results_total{type="accepted"} 91.000000
# TYPE xmrstak_difficulty_bucket gauge
xmrstak_difficulty_bucket 150.000000
# TYPE xmrstak_pool_connection_ping_seconds gauge
xmrstak_pool_connection_ping_seconds{address="xxx:3333"} 42.000000
# TYPE xmrstak_hash_rate_bytes_total gauge
xmrstak_hash_rate_bytes_total{le="900"} Nan
xmrstak_hash_rate_bytes_total{le="10"} 252.892318
xmrstak_hash_rate_bytes_total{le="60"} Nan
# TYPE xmrstak_hash_rate_bytes gauge
xmrstak_hash_rate_bytes{device="cpu",le="10",thread_id="12"} 18.3234
xmrstak_hash_rate_bytes{device="cpu",le="60",thread_id="12"} 17.378082
xmrstak_hash_rate_bytes{device="cpu",le="900",thread_id="5"} 21.585160

// CC: @yorik

yorik

Thank you! I left some comments, but the main question: why to export rates, when just total would be enough?

yorik · 2018-03-21T12:03:30Z

xmrstak/http/httpd.cpp

 		rsp = MHD_create_response_from_buffer(str.size(), (void*)str.c_str(), MHD_RESPMEM_MUST_COPY);
 		MHD_add_response_header(rsp, "Content-Type", "application/json; charset=utf-8");
 	}
+	else if(strcasecmp(url, "/prometheus") == 0)


Why no to go with default "/metrics" ?

Because it's Prometheus's related format of output. I don't have problem with metric name too. Name can be changed.

It would require less custom settings, so I think it's better, but this is minor.

yorik · 2018-03-21T12:08:37Z

xmrstak/misc/executor.cpp

+	sMetric.gPoolPing = &gauge_family_pool_ping.Add({{"address", "xxx:3333"}});
+
+	auto& gauge_family_hr_total = BuildGauge()
+		.Name("xmrstak_hash_rate_bytes_total")


I would named it "xmrstak_hash_rate_total"

By Prometheus convention units should be in name.

But unit isn't bytes in this case, it's hashes per second, or am I missing something?

yorik · 2018-03-21T12:09:02Z

xmrstak/misc/executor.cpp

+	sMetric.agHashRateTotal[2] = &gauge_family_hr_total.Add({{"le", "900"}});
+
+	auto& gauge_family_hr = BuildGauge()
+		.Name("xmrstak_hash_rate_bytes")


I would remove _bytes from here the metric name. Also it would be nice to have device and thread_id labels here too.

Bytes represent units: https://prometheus.io/docs/practices/naming/. Labels already exits, look at 1348-1362 there are dynamically created.

Missed that. Thanks!

t0kieu · 2018-03-21T15:00:17Z

This is optimization :-) If you want total hash rate and you have big amount of data, then query to Prometheus backed is more performed, instead of doing sum calculation on all data aggregated by thread...

xmrstak_hash_rate_bytes_total represents summary of all hash rates from all threads and all devices.
xmrstak_hash_rate_bytes represents specific hash rate for specific thread and device identified via labels.

psychocrypt · 2018-03-24T20:54:32Z

THX for the PR, I will check this PR after the next release. I need first to check what prometheus

DrStein99 · 2018-03-25T20:06:48Z

Hi. I am interested in this, and was wondering if there was a link or document anywhere that can help me set this up (outside of my own obvious searches by using google).

I believe it is referencing prometheus.h somewhere, and not sure if the data collection software has to be installed on the xmr-stak miner, or what steps I need to try and set it up.

erikgajdos · 2018-03-26T20:44:58Z

Hey, looks good, looking forward once it will be merged and released. Any ETA?

t0kieu · 2018-03-26T22:57:53Z

@DrStein99 Try that simple steps:

make build xmr-stak with promethes support
run with http server enabled for example on port 12000
download https://prometheus.io/download/ & run it.
use that simple configuration for probmethes prometheus.yml (adjust IP:PORT of your miner).

global:
  scrape_interval:     15s # By default, scrape targets every 15 seconds.
scrape_configs:
  # XMR-STAK application
  - job_name: xmr-stak-application
    metrics_path: "/prometheus"
    static_configs:
      - targets: ['192.168.1.2:12000']
        labels:
          instance: your-hostname.com
          group: intranet

After a few seconds, promethes will be periodically connect (in that case every 15s) to your xmr-stak miner and it will be collects measured data.

Tualua · 2019-10-12T01:10:32Z

Any change to merge it into a current version?

psychocrypt · 2019-10-12T07:56:08Z

currently I have no time. The changes very old and must be verified/tested first.

Implement basic metric with prometheus

1bf4f37

yorik reviewed Mar 21, 2018

View reviewed changes

psychocrypt self-assigned this Mar 24, 2018

psychocrypt added the feature label Mar 24, 2018

Merge branch 'dev' of https://github.com/fireice-uk/xmr-stak into dev

c55b5c3

Conversation

t0kieu commented Mar 21, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

yorik left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

t0kieu commented Mar 21, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

psychocrypt commented Mar 24, 2018

Uh oh!

DrStein99 commented Mar 25, 2018

Uh oh!

erikgajdos commented Mar 26, 2018

Uh oh!

t0kieu commented Mar 26, 2018

Uh oh!

Tualua commented Oct 12, 2019

Uh oh!

psychocrypt commented Oct 12, 2019 via email

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

t0kieu commented Mar 21, 2018 •

edited

Loading

t0kieu commented Mar 21, 2018 •

edited

Loading