Port Statistics Dashboard Not Working

[jhughesbiot]

The port statistics dashboard appears to be broken. I have cloned a clean copy of this repo, changed the faucet.yml config to match my setup, configured grafana to talk to the prometheus data service and setup all dashboards accordingly. The other two dashboards show information as expected but the port statistics do not show any datapath names. I can manually enter them and it resolves the port list but no data is ever resolved.

I am running this via docker-compose so all the versions there are the versions being used. Is this a broken dashboard? Prometheus version issue? Some other compatibility issue with the dashboard itself?

[courtland]

I tried the latest port stats dashboard from this repo and am not having any issues, however it could just be something different with my setup.

If you query of_port_rx_bytes or a similar metric via Grafana Explore does it show any data?

Is gauge running and configured correctly? I think it should be with the default compose config, but that's what populates the port metrics.

I'm not a maintainer, so perhaps someone else can chime in, but it may be helpful to see your modified faucet and gauge config files.

[jhughesbiot]

Could you elaborate on what the test using the grafana explorer looks like? Not really sure what I would have to do to test this.

I'm using the docker-compose file from this repository that sets up all services so I see gauge running in docker. The gauge.yaml file is the one from the /etc/ folder in this repository. https://github.com/faucetsdn/faucet/blob/master/etc/faucet/gauge.yaml

I can also see that the faucet and gauge controller are both connected to the ovs bridge so not really sure what the issue is.

709f180d-9301-4da3-b93c-4e8f93e94763
    Bridge ovs-sw
        Controller "tcp:192.168.10.2:6654"
            is_connected: true
        Controller "tcp:192.168.10.2:6653"
            is_connected: true
        Port ens10
            Interface ens10
        Port dhcp-eth1
            Interface dhcp-eth1
        Port ovs-sw
            Interface ovs-sw
                type: internal
    ovs_version: "2.13.8"

[courtland]

Could you elaborate on what the test using the grafana explorer looks like? Not really sure what I would have to do to test this.

In your Prometheus database you should have an of_port_rx_bytes metric, among others populated by Gauge. You can check this via the Prometheus or Grafana UIs. IMO Grafana Explore is easier to look at metrics. If the metrics don't exist then the problem lies outside the dashboard.

[flogriesser]

I had the issue that after a fresh install the of_port_rx_bytes was missing in grafana.
Somehow a reboot fixed the issue.
(Seen on debian 12 and ubuntu 22.04)

[joaopedropsilva]

Any updates on this issue? I am trying to query of_port_tx_dropped metric, but it seems that it is not visible to prometheus. I am running faucet on a docker container, and after consulting /metrics endpoint from gauge service I can see the metric I am willing to use, but it is commented, I don't know if it impacts its accessibility. I am using the default gauge.yaml configuration available at the docs.

If I try to seek the metric on prometheus metrics explorer I get no results:

[gizmoguy]

@joaopedropsilva do you actually have a datapath connected to gauge? metrics won't show up until the datapath is connected.

You can check like this (or check gauge log file):

$ curl localhost:9303 2>/dev/null | grep dp_status
# HELP dp_status status of datapaths
# TYPE dp_status gauge
dp_status{dp_id="0x1",dp_name="dp1"} 1.0

[joaopedropsilva]

@gizmoguy after running the above command and also checking gauge logs I have noticed that my datapath is not actually connected to gauge. Even though I still did not manage to establish that connection. I am not sure if I have misconfigured gauge, I have also tried to explicitly set my datapath name in gauge.yaml in port_status_poller instead of using all_dps: true, but it did not work as well.

Command output:

Gauge logs:

Faucet logs:

gauge.yaml:

[gizmoguy]

It's more likely your datapath is misconfigured rather than gauge, I don't see any datapaths attempting to connect to gauge in your log screenshot.

I recommend you follow the getting started faucet tutorial which shows you how to set all this up: https://docs.faucet.nz/en/latest/tutorials/first_time.html

[joaopedropsilva]

@gizmoguy my datapath was indeed misconfigured. I was having trouble to assign gauge as a controller for my virtual network and after retrying all the steps it did work, I could see the metrics I was looking for. Thanks for the help!