Add/expose metrics only specific ns #94
Conversation
…for all ns by default, unless is set to true
…for all ns by default, unless is set to true
|
Hi @kimiyasharifi, thanks for you pull request. Overall I am for this feature, but we do need to revert of some changes. I will annotate those now. I also enabled the e2e test on this PR incase you can't run the test local. Currently it's failing due to the |
| select { | ||
| case <-ticker.C: |
There was a problem hiding this comment.
What is the benefit of doing it this way?
| @@ -139,12 +139,28 @@ func (n *Node) Watch() { | |||
| // Start the informer to begin watching for Node events | |||
| go sharedInformerFactory.Start(stopCh) | |||
|
|
|||
| // Use a ticker for periodic actions | |||
| ticker := time.NewTicker(time.Duration(n.sampleInterval) * time.Second) | |||
There was a problem hiding this comment.
Why should we do it this way?
| @@ -23,6 +27,35 @@ var ( | |||
| Pod pod.Collector | |||
| ) | |||
|
|
|||
| func getMonitoredNamespaces(clientset *kubernetes.Clientset) ([]string, error) { | |||
There was a problem hiding this comment.
Close. This will get our namespaces on init only. We will want to update our cache of namespaces periodically incase need ones were added.
A good example would be this node watch with a shard informer.
https://github.com/jmcgrath207/k8s-ephemeral-storage-metrics/blob/master/pkg/node/k8s.go#L106
Wouldn't be against following the same pattern and putting that code under pkg/k8s.go
|
Hey not see any activity. Closing feel free to reopen once changes are completed. |
This pull request introduces changes to enable the exporter to expose metrics only for ns labeled with specific label. Introduced a new env var
EPHEMERAL_STORAGE_LABELto specify the label selector for ns to monitor.getMonitoredNamespaces(clientset *kubernetes.Clientset) ([]string, error)now reads the label selector from theEPHEMERAL_STORAGE_LABELenv var.setMetrics(clientset *kubernetes.Clientset, nodeName string, monitoredNamespaces map[string]bool)now receives a map of monitored ns to check if a pod’s ns is in the monitored list.Additionally, this pull request adds the capability to expose metrics for all namespaces by default, unless a specific environment variable,
ENABLE_SPECIFIC_NAMESPACE_MONITORING, is set totrue. This logic is implemented in thegetMonitoredNamespacesfunction.