Beginner Issues

[mgg1010]

Hi

First, I'm seeing a lot of go-related metrics in the output

go_gc_duration_seconds{quantile="0"} 3.2688e-05

I'm new to prometheus, but doesn't spamming it with 100s of extra metrics cause
some memory use - or should I configure prom to ignore anything starting with 'go'

Am I missing something?

Second, I'm sending messages like this:

topic: watchdog/esp-garage:
Payload: {"message": "Alive"}

I get this on the output:

received_messages{status="storeError",topic="watchdog/esp-garage"} 34

previously I sent non-JSON messages and it was happier, but I didn't see
a direct metric for my line.

My configuration is below - any thoughts?

 topic_path: watchdog/#
 device_id_regex: "(.*/)?(?P<deviceid>.*)"
 qos: 0
cache:
   timeout: 24h
json_parsing:
 - prom_name: watchdog
   mqtt_name: message
   help: Watchdog message
   type: gauge

Thanks

Martin Green

[wmoss]

Regarding the error, I think you want put the specific message configuration under metrics: not json_parsing:. If would be helpful if you posted the logs from the service though, as that would have more information.

Regarding the go... metrics, I was actually noticing the same thing @hikhvar. Thoughts on adding a configuration flag to remove those?

[hikhvar]

Hey,
it's quite normal in the prometheus eco system to have those runtime specific go_... metrics, and I don't see any problems in having them. The canonical mechansim to drop them is via relabeling in the prometheus server: https://medium.com/quiq-blog/prometheus-relabeling-tricks-6ae62c56cbda

Regarding your initial problem:
The payload you presented, don't have a number as value. mqtt2prometheus assumes by default, that the values are numbers, because prometheus metrics can only have numeric values. If you want to map a string like here:

{"message": "Alive"}

you must define a mapping. Example:

 topic_path: watchdog/#
 device_id_regex: "(.*/)?(?P<deviceid>.*)"
 qos: 0
cache:
   timeout: 24h
json_parsing:
 - prom_name: watchdog
   mqtt_name: message
   help: Watchdog message
   type: gauge
   string_value_mapping:
     map:
       Alive: 1
       Dead: 0
      # Metric value to use if a match cannot be found in the map above.
      # If not specified, parsing error will occur.
     error_value: -1

[wmoss]

Yeah, I ended up doing something similar in my prometheus config and I guess that's just as good a solution as adding a flag, probably easier. If it's helpful, my config looks like,

scrape_configs:
  - job_name: mqtt2prometheus
    scrape_interval: 10s
    static_configs:
      - targets:
        - "localhost:9641"
    metric_relabel_configs:
      - source_labels: ["__name__"]
        regex: <metrics to keep here>
        action: keep

@mgg1010 seems like it wouldn't be a terrible entry in the FAQ of the main Readme if you wanted to add the answer to your question there.