I am running the Alloy diagnostic dashboard. It is in Microk8s and I’m installing the entire LGTM stack with Helm. Alloy installs. Everything is reading as green. I can see logs in Grafana so I know that Loki is getting sent data. And yet, when I go to http://127.0.0.1:12345/ through a port-forward, the data panel is blank. There is nothing in the network tab of dev tools. Nothing in the console. Everything is looking good. Further adding to my confusion. I have two systems: an AMD desktop and an Intel laptop. Both systems are running the same cluster init Powershell script that I created. The dashboard fails to load on my AMD system but works just fine on my Intel laptop.
127.0.0.1 is the loopback or home address, so it will work on the local machine but not on the other.
use the actual ip address of the other machine and see if that works?
Sorry. I should have specified that I am using kubectl’s port-forward to route my local IP to the alloy service. E.g. kubectl port-forward service/alloy -n observability 12345:12345
This was working on both of my systems up until a few days ago when suddenly my desktop started doing this. The preceding event was me configuring a bunch of new Alloy components. I spent two days stepping through them to see which one was causing this only to discover that going back to my previous, known-good config displayed the same behavior. And as I mentioned, this works just fine on my laptop.
can you hit port 12345 on both machines from a browser?
Yep. My desktop gets the dashboard but does not receive any of the component data. Meanwhile my laptop gets everything successfully. I can look in devtools and see all of the same requests on my laptop as on my desktop, but the components request gets a JSON payload on my laptop while it gets nothing on my desktop. The response code is 200, though, indicating that Alloy thinks everything is fine.
can you compare your two yaml files to see if both are valid?
what do you have configured for the listen address on the machien that is not working?
–server.http.listen-addr=LISTEN_ADDR:12345
The Alloy configs are completely identical between my two systems. It is the same repo, actually. It does not seem to matter how simple I make the config. I have reduced it to a single discovery component, but the components still fail to load in the UI.
As for the listen address, I don’t configure via flags like that. I’m using the Helm chart with the default values for that, which is 12345.
You can see the values here: https://artifacthub.io/packages/helm/grafana/alloy?modal=values&path=alloy.listenAddr