PMM-Server not collecting own metrics

[AND]

Hello.

I’ve instaled a brand new pmm-server following the steps for a docker deploy. But I cant see the pmm-server metricss. Just a error “Unknown error during query transaction. Please check JS console logs.” in left-top of dashboards. Its a GCP Compute Machine.
What do I miss?

[Michael_Coburn]

Hi @AND thank you for posting to the Percona forums!
I’ve not yet seen that case where PMM Server isn’t able to display its own metrics. Can you see metrics from other hosts you have configured?
Please share with us the logs from the system - you can access them via the PMM menu at the top right and going to PMM Settings → Download logs.
Or you can get a shell inside the container and inspect the logs directly:

docker exec -it pmm-server bash
cd /srv/logs

In here I would start with grafana.log, or just tail them all and try to reload the dashboard to see what entries of interest are logged. Please share here and we’ll try to help you out!

[AND]

Checked all logfiles and dont found any error message logs.
grafana.log just have commom grafana infor about restart.
But one entry looks awkard for me:
t=2021-05-07T16:54:56+0000 lvl=info msg="HTTP Server Listen" logger=http.server address=[::]:3000 protocol=http subUrl=/graph socket=
addres dont have any ip?

I also have registered a node in this machine, and also dont have received any metrics.

[Michael_Coburn]

Hi @AND
Is your server enabled for IPv6 only? i.e. you haven’t intentionally disabled IPv4?
Since you have registered nodes but aren’t seeing any graphs being populated, can you share then the contents of the client logs? These will be logged to syslog.

[AND]

you haven’t intentionally disabled IPv4?

No

I have created another pmm-server container, the only data that shows on dashboard is system-uptime and ram.

Do I need any firewall rules, even for get “local” metrics?

These will be logged to syslog.

I dont found, could you tell me please where?

Edit:
I found this in my browser console

[Michael_Coburn]

Hi @AND

1. You do not need any firewall rules to be additionally configured for local metrics - these are actually scraped from within the pmm-server container over localhost
2. Regarding syslog - if you are using Ubuntu then look in /var/log/syslog and with RHEL look in /var/log/messages

I just deployed a GCompute Ubuntu 20.04 instance and I have it working correctly. I had to define additional inbound firewall rules (443/tcp) in the Console but otherwise the instance operates as expected. You can check it out here using admin / admin credentials:
https://34.66.223.31/graph/d/node-instance-summary/node-summary?orgId=1&from=now-15m&to=now&var-interval=$__auto_interval_interval&var-region=All&var-node_type=All&var-node_name=All&var-service_name=All&var-database=All&var-service_type=All&var-environment=All&var-cluster=All&var-replication_set=All&var-username=All&var-schema=All&refresh=1m

My suggestion for you is to deploy another instance and see if you can reproduce.

[AND]

I’ve created a brand new project with all/all rules and it work fine and then it works fine.

But I cant imagine something that was bloking my pmm inside this project.
I also created a new pmm-server and at end of grafana.log get this:

[AND]

We finished,

First it was a problem with the docker user, some problem with the docker configuration did not create it. Creating and making the “post-installation” docker solved a problem.

Another was a firewall rule for VPC traffic, which was not allowed, we have to allow the machine to reach them.