Description
Description
We already have Grafana
running in Kubernetes
cluster with horizontal auto scale feature enabled so Grafana pods can increase or decrease the amount of replicas based on usage. These Grafana
pods are pointing to external postgres
database running as a service in AWS RDS.
Is there a way to run PMM
separated from Grafana
?
Suggested solution
I was thinking in a feature flag based on environments variables like describe in HA document:
Ex.:
- For external
postgres
:PG_EXTERNAL_DB = true | false
,PG_HOST = external.postgres.local
,PG_USER = pg-username
,PG_PASSWORD = some-random-p4ssw0rd
- For external
grafana
:GF_EXTERNAL = true | false
- For external
victoria metrics
:VM_EXTERNAL = true | false
,VM_HOST = external.victoria-metrics.local
- And so on ...
In the running Grafana
we can install or copy plugins and import PMM dashboards before running the PMM pod for the first time.
In the PMM pod we can set autostart = false
inside the *.ini
(/etc/supervisord.d/grafana.ini
, /etc/supervisord.d/pmm.ini
, ...) supervisord files according to the service we are running separatedly (postgres, clickhouse, victoria metrics, ...).
We can also set any config file based on the same environment variables to point to external services.
Any suggestion are welcome.
Additional context
I was taking a look at the HA doc and I recommend to change the usage of ip addresses
to FQDN
from each external service to avoid the necessity of fetch new ip addresses, change config files and restart PMM in each change. This way we can run it inside a kubernetes cluster.
Code of Conduct
- I agree to follow this project's Code of Conduct