deno-observability

Metrics exporting for Deno’s runtime, initially targetting OpenMetrics (Prometheus) and the Datadog API

Project Status

The code in this repository will be a sorta scratch-pad until a v1.0 possibly eventually happens. I’m trying to reconsile the design differences between pushing & polling, and figure out one way of instrumenting metrics that will be able to transmit to different types of metric sinks.

The current version requires Deno 1.15 or later.

The Deno --unstable flag is not required, while any latency metrics will be very rough unless --allow-hrtime is passed.

Prometheus Example

It’s standard to run a Prometheus exporter on its own dedicated port to keep it separate from whatever API you may be hosting. Or if you aren’t hosting an API at all, you may not already have any HTTP server.

Here’s an example of that:

import { runMetricsServer } from "https://deno.land/x/observability@v0.1.0/sinks/openmetrics/server.ts";

runMetricsServer({ port: 9090 });
console.log("Now serving http://localhost:9090/metrics");

You can also run the demo directly, which simulates extra HTTP traffic to make the metrics more full:

deno run --allow-net https://deno.land/x/observability@v0.1.0/demo.ts

If you would like to expose metrics on your existing web server, you’ll want to look at mod.ts for inspiration instead of importing it as-is.

Deno Metrics

• deno_ops_dispatched: # of operations started, split into 3 categories & faceted by op name

• deno_ops_completed: # of operations finished, split into 3 categories & faceted by op name

• deno_open_resources: # of currently registered Deno resources, split by resource type.

  • A process starts up with 3: stdin, stdout, stderr.
  • Starting the metrics server will add 1 tcpListener.
  • Your metrics HTTP request seemingly always shows up as an additional tcpStream.
  • All other resources are from your application’s own code (or libraries in use).

• deno_memory_rss_bytes: # of total resident bytes held by the Deno process

• deno_memory_heap_total_bytes: allocated size of the Deno heap space

• deno_memory_heap_used_bytes: used size of the Deno heap space

• deno_memory_external_bytes: Not Yet Implemented in Deno. Currently 0.

Here’s an example of the payload:

# TYPE deno_ops_dispatched counter
deno_ops_dispatched_total{op_type="sync"} 179
deno_ops_dispatched_total{op_type="async"} 252
# TYPE deno_ops_completed counter
deno_ops_completed_total{op_type="sync"} 179
deno_ops_completed_total{op_type="async"} 245
# TYPE deno_open_resources gauge
deno_open_resources{res_type="stdin"} 1
deno_open_resources{res_type="stdout"} 1
deno_open_resources{res_type="stderr"} 1
deno_open_resources{res_type="tcpListener"} 1
deno_open_resources{res_type="child"} 30
deno_open_resources{res_type="tcpStream"} 2
# TYPE deno_memory_rss_bytes gauge
# UNIT deno_memory_rss_bytes bytes
deno_memory_rss_bytes 3232380
# TYPE deno_memory_heap_total_bytes gauge
# UNIT deno_memory_heap_total_bytes bytes
deno_memory_heap_total_bytes 3948544
# TYPE deno_memory_heap_used_bytes gauge
# UNIT deno_memory_heap_used_bytes bytes
deno_memory_heap_used_bytes 3690960
# TYPE deno_memory_external_bytes gauge
# UNIT deno_memory_external_bytes bytes
deno_memory_external_bytes 16384

HTTP Server Metrics

• denohttp_handled_requests: Number of HTTP requests that have been received and responded to in full.
• denohttp_requests_in_flight: Current number of HTTP requests being served.
• denohttp_request_duration_seconds: A histogram of the HTTP request durations, including writing a response.
• denohttp_response_bytes_total: Number of bytes transmitted in response to HTTP requests. Includes a facet for which aspect of the HTTP response the bytes were a part of.

# TYPE denohttp_handled_requests counter
# HELP denohttp_handled_requests Number of HTTP requests that have been received and responded to in full.
denohttp_handled_requests{code="404",method="get"} 1218
denohttp_handled_requests{code="200",method="get"} 6
# TYPE denohttp_requests_in_flight gauge
# HELP denohttp_requests_in_flight Current number of HTTP requests being served.
denohttp_requests_in_flight 1
# TYPE denohttp_request_duration_seconds histogram
# UNIT denohttp_request_duration_seconds seconds
# HELP denohttp_request_duration_seconds A histogram of the HTTP request durations, including writing a response.
denohttp_request_duration_seconds_bucket{le="0.01"} 1219
denohttp_request_duration_seconds_bucket{le="0.1"} 1224
denohttp_request_duration_seconds_bucket{le="1"} 1224
denohttp_request_duration_seconds_bucket{le="+Inf"} 1224
denohttp_request_duration_seconds_sum 0.704
denohttp_request_duration_seconds_count 1224
# TYPE denohttp_response_bytes counter
# UNIT denohttp_response_bytes bytes
# HELP denohttp_response_bytes Number of bytes transmitted in response to HTTP requests.
denohttp_response_bytes_total{purpose="body"} 25247
denohttp_response_bytes_total{purpose="framing"} 31782
denohttp_response_bytes_total{purpose="header"} 24936

TODO

• Serve data from Deno.metrics()
• Serve data from Deno.resources()
• Accept custom metrics from the user’s code
• Include ‘created’ timestamp on counters to aid in monotonic tracking
• Publish on deno.land
• Keep track of previous metrics to include zero gauges in future bodies

License

MIT