How to disable auto-detection for Prometheus?

[SpraxDev]

I am seeing duplicate Prometheus exporter metrics in my dashboard and can see netdata trying to add/connect to another service that just happens to use a port from the very long list of known Prometheus exporter ports (port 56300 is actually not on it).

When I remove my explicit 'TeamSpeak' configuration from the config, the metrics with the ugly name remain and continue to receive new data, while the 'TeamSpeak' one stops receiving data (as expected).

I wasn't able to find anything related to Prometheus auto-detection in the logs.

Some reasons for me to configure the exporters manually:

• I want to give them a proper and nice name
• I might want to add filters etc.
• I don't mind manually telling netdata what endpoints it should check, in fact I would prefer to be able to do that

My go.d/prometheus.conf:

# https://github.com/netdata/netdata/tree/master/src/go/plugin/go.d/collector/prometheus#readme

autodetection_retry: 0
jobs:
 - name: nftables
   url: 'http://[::1]:56300/metrics'
 - name: TeamSpeak
   url: 'http://127.0.0.1:56301/metrics'
   selector:
     deny:
       - promhttp_metric_handler_*
 - name: DNSCrypt-Proxy
   url: 'http://[::1]:57014/metrics'

My logs filtered by Prometheus (journalctl _SYSTEMD_INVOCATION_ID="$(systemctl show --value --property=InvocationID netdata)" --namespace=netdata --grep prometheus):

level=info msg="check success" plugin=go.d collector=prometheus job=nftables
level=info msg="started, data collection interval 10s" plugin=go.d collector=prometheus job=nftables
level=info msg="check success" plugin=go.d collector=prometheus job=DNSCrypt-Proxy
level=info msg="started, data collection interval 10s" plugin=go.d collector=prometheus job=DNSCrypt-Proxy
level=info msg="check success" plugin=go.d collector=prometheus job=exporter_56301_local
level=info msg="started, data collection interval 10s" plugin=go.d collector=prometheus job=exporter_56301_local
level=info msg="check success" plugin=go.d collector=prometheus job=exporter_56300_local
level=info msg="started, data collection interval 10s" plugin=go.d collector=prometheus job=exporter_56300_local
level=error msg="check failed: Get \"http://127.0.0.1:9987/metrics\": dial tcp 127.0.0.1:9987: connect: connection refused" plugin=go.d collector=prometheus job=netapp_solidfire_exporter_local

Thanks for any tips or insight ^^

[ilyam8]

Hi, @SpraxDev. You can disable autodetection by configuring go.d/sd/net_listeners.conf

╭─ilyam@ilyam-deb13-work [📁 netdata] [12:35:31]
╰─❯ pwd
/opt/netdata/etc/netdata

╭─ilyam@ilyam-deb13-work [📁 netdata] [12:35:34]
╰─❯ sudo ./edit-config go.d/sd/net_listeners.conf

• Set disabled to yes
• Restart netdata

[ilyam8]

port 56300 is actually not on it

Netdata detects exporters not only by known Prometheus ports, but also by the binary name - if the binary name contains "exporter", then it is most likely a Prometheus exporter.

[SpraxDev]

Thanks for pointing me to the relevant file! And now I understand why these exporters 'suddenly' showed up. I was already confused whether I got netdata in some weird state or broke a configuration file.
#20348

I'd say non-Prometheus exporters can also habe exporter in their (binary) name but being able to disable it makes me happy already.
Although I only want to disable the Prometheus detection...

Looks like the file is responsible for auto-detection in general, so I will try to remove the two Prometheus related sections and see where that leads me.