client_python questions about multiprocess

[lamp...@gmail.com]

Hello,

I am trying to integrate metrics into a multiprocess application. It is not HTTP (gunicorn) based so I am a little lost trying to follow such examples. Basically, I have a master process which spins up a series of long-lived child processes which perform various functions. I am trying to use the mater process to do the exposition like so:

registry = CollectorRegistry()

multiprocess.MultiProcessCollector(registry)

start_http_server(8000, registry=registry)

In a child process I have a simple Counter to test with:

c_logs_ingested = Counter('logs_listener_logs_ingested', 'Count of ingested log messages')

c_logs_ingested.labels(listener_type=self._listener_type, address=self.address, port=self.port).inc()

I have set the collection dir environment variable:

export prometheus_multiproc_dir=/tmp/prom

And when running, I can see that directory being populated:

$ ls -la

total 64

drwxrwxr-x  2 andersonjd andersonjd    4096 May  8 12:57 .

drwxrwxrwt 44 root       root         49152 May  8 12:57 ..

-rw-rw-r--  1 andersonjd andersonjd 1048576 May  8 12:45 counter_23714.db

-rw-rw-r--  1 andersonjd andersonjd 1048576 May  8 12:45 counter_23751.db

-rw-rw-r--  1 andersonjd andersonjd 1048576 May  8 12:57 counter_24807.db

But when I look at the metrics being exported at localhost:8000, it only contains the default environment metrics. What am I missing here?

Thank you for your time,

John

[Brian Brazil]

Are you sure the environment variable is set for the master? 

--

Brian Brazil

www.robustperception.io

[lamp...@gmail.com]

Yep, that was it. Thank's for the rubber duck sanity check :)