Monitoring Insanic¶

Insanic provides several default endpoints when you run your application. These are to help with general health checks, metrics, and to ping intra service communications.

Health Endpoint¶

The health endpoint is served from /<your_service_name>/health/. This provides basic information about the service that is running.

If we have this application…

from insanic import Insanic
from insanic.conf import settings

__version__ = '0.1.1'

settings.configure()
app = Insanic('example', version=__version__)

And if we run and curl…

$ curl http://localhost:8000/example/health/
{
    "service":"example",
    "application_version":"0.1.0",
    "status":"OK",
    "insanic_version":"0.8.4.dev0",
    "ip":"172.20.10.10"
} # formatted for readability

Metrics Endpoint¶

The metrics endpoint is served from /<your service name>/metrics/. This endpoint is provide basic metrics of the application and machine/container that it is running in. This is because transparent application metrics are very important in a distributed system.

This endpoint provides information for the following metrics.

Total asyncio Task Count
Active asyncio Task Count
Memory Usage/Percentage
CPU Usage
Processed Request Count

And the endpoint provides these metrics in 2 formats.

JSON
Prometheus

With our example application from above…

$ curl http://localhost:8000/example/metrics/
# HELP python_gc_collected_objects Objects collected during gc
# TYPE python_gc_collected_objects histogram
python_gc_collected_objects_bucket{generation="0",le="500.0"} 16.0

... truncated for brevity

# HELP service_info Meta data about this instance.
# TYPE service_info gauge
service_info{application_version="0.1.0",insanic_version="0.8.4.dev0",ip="172.20.10.10",service="example",status="OK"} 1.0

$ curl http://localhost:8000/example/metrics/?json
{
    "total_task_count":1,
    "active_task_count":1,
    "request_count":6.0,
    "proc_rss_mem_bytes":13205504.0,
    "proc_rss_mem_perc":0.15382766723632812,
    "proc_cpu_perc":0.0,
    "timestamp":1600700128.8033018
} # formatted for readability

Ping Endpoint¶

The ping endpoint is served from /<your service name>/ping/.

This endpoint pings the service and provides request process duration metrics for the respective service (usually 0ms). However, its real power lies in its ability to ping other services.

A call to the endpoint gathers the services defined in the SERVICE_CONNECTIONS and REQUIRED_SERVICE_CONNECTIONS settings and also sends a ping request to all of them. Depth can be set to determine how far in the mesh you want to traverse with the depth query parameter.

This could be useful for creating a trace diagram of which service talks to who, if you have some sort of tracing stack.

Warning

Requests with a large depth value should be avoided in a production environment as it could quickly flood the network. Especially if you have circular connections.

Again, with our example application above…

$ curl http://localhost:8000/example/ping/
{"response":"pong","process_time":"0 ms"}
$ curl http://localhost:8000/example/ping/?depth=1

If you don’t need these endpoints¶

If you don’t need these, you can turn them off by sending an argument to Insanic on initialization.

app = Insanic(
    "nomonitor",
    version='0.0.0',
    attach_monitor_endpoints=False
)