Support for multiple k8s clusters

9strands commented 5 months ago

So far the initial service looks very interesting and I'm tinkering with it a bit.

However, I think most users are going to wind up needing to monitor multiple clusters - a dev one, a test one, and probably 2 production clusters. I can't figure out how this can currently work with it - since it needs a specific KUBERNETES_MASTER env var defined to operate.

My guess is that this would also need some changes to the SQL schema, potentially to support with potentially multiple agents running on different nodes polling different clusters - but updating the same backend DB.

ngoeddel-openi commented 4 months ago

Oh, I was just browsing through the issues and saw this one. We also want to have one IcingaWeb and multiple Kubernetes clusters to be included in the UI. It seems there is no way right now to somehow workaround that. I hope this can be added soon. I really would appreciate that.

9strands commented 3 months ago

I run about 18 K8s clusters, and would love to use Icinga-k8s in order to implement the monitoring (using these: https://github.com/redhat-cop/rhdp-monitoring-scripts/tree/main/openshift/ mostly)

Right now I have custom scripts reaching out from Icinga to each cluster, but having a full-and-proper monitoring solution for each cluster would make my life infinitely easier from a monitoring perspective.

PeterLustig1337 commented 2 months ago

I also got Issues running icinga-kubernetes to monitor a cluster. The systemd service is unable to complete start somehow and it always times out. API request are all being made and I got all cluster information on the web module, but when it reaches the timeout for service start it just exits. And everytime this happens the database is somehow corrupt and I need to drop the database and recreate it. Otherwise i get a lot of "connection reset by peer" errors on the next start.

lokidaibel commented 1 month ago

I also got Issues running icinga-kubernetes to monitor a cluster. The systemd service is unable to complete start somehow and it always times out. API request are all being made and I got all cluster information on the web module, but when it reaches the timeout for service start it just exits. And everytime this happens the database is somehow corrupt and I need to drop the database and recreate it. Otherwise i get a lot of "connection reset by peer" errors on the next start.

Do you try V0.2 Had the same Issues with v0.1.

I would also like multi Cluster Support ! PUSH !

sebold-minova commented 1 month ago

Also interested in this

lippserd commented 3 weeks ago

@jhoxhaa implementation for the daemon side should be roughly as follows:

Create a cluster table with a single column uuid.
Include a cluster_uuid column in all relevant tables.
Generate a cluster's uuid using the uid of the kube-system namespace.
In Icinga DB, we have a similar concept called environments. We utilize context.WithValue() to pass the environment^1 and ScopedEntity to select entities within the current scope^2. Apply the same approach here.
Add ClusterUuid to schemav1#Meta and find a good place where to actually set it.

PeterLustig1337 commented 3 weeks ago

I also got Issues running icinga-kubernetes to monitor a cluster. The systemd service is unable to complete start somehow and it always times out. API request are all being made and I got all cluster information on the web module, but when it reaches the timeout for service start it just exits. And everytime this happens the database is somehow corrupt and I need to drop the database and recreate it. Otherwise i get a lot of "connection reset by peer" errors on the next start.

Do you try V0.2 Had the same Issues with v0.1.

I would also like multi Cluster Support ! PUSH !

yes, got the issue with v0.2

Icinga / icinga-kubernetes

Support for multiple k8s clusters #95