[Prometheus] Add ray_clusters_created_total metric
#3204
Conversation
win5923
changed the title
ray_clusters_created_total metric
|
@troychiu PTAL, thx |
win5923
force-pushed
the
metrics/ray_clusters_created_total
branch
7 times, most recently
from
9976c1b to
92e7b83
Compare
win5923
force-pushed
the
metrics/ray_clusters_created_total
branch
4 times, most recently
from
c446b4a to
6fa17fc
Compare
| } else if len(headPods.Items) == 0 { | ||
| // Create head Pod if it does not exist. | ||
| logger.Info("reconcilePods: Found 0 head Pods; creating a head Pod for the RayCluster.") | ||
|
|
There was a problem hiding this comment.
Maybe we should increment after successfully create the head pod to avoid incrementing multiple times?
There was a problem hiding this comment.
It's still possible to be counted multiple times. KubeRay is possible to delete the head Pod if the restart policy is Never.
| // Create head Pod if it does not exist. | ||
| logger.Info("reconcilePods: Found 0 head Pods; creating a head Pod for the RayCluster.") | ||
|
|
||
| if r.EnableMetrics { |
There was a problem hiding this comment.
Is it possible to have a function in metrics.go handle EnableMetrics in a general way?
|
|
||
| if r.EnableMetrics { | ||
| creatorCRDType := getCreatorCRDType(*instance) | ||
| // Increment the counter for ray_clusters_created_total metric based on the creator CRD type. |
There was a problem hiding this comment.
maybe we can move these checkings to ray_cluster_metrics.go to simplify the operator's logic?
There was a problem hiding this comment.
or maybe still in this file but move to a function? What do you think it's better?
There was a problem hiding this comment.
I added a function called emitRayClusterMetrics in raycluster_controller.go. It takes a metric name and decides which metric to emit based on that.
WDYT?
win5923
force-pushed
the
metrics/ray_clusters_created_total
branch
2 times, most recently
from
a4290f9 to
8baf776
Compare
win5923
force-pushed
the
metrics/ray_clusters_created_total
branch
from
8baf776 to
ef1a740
Compare
| "sigs.k8s.io/controller-runtime/pkg/metrics" | ||
| ) | ||
|
|
||
| var rayClustersCreatedCounter = prometheus.NewCounterVec( |
There was a problem hiding this comment.
It's better not to use package-level variable. Can you open another PR first to improve it by:
ray_cluster_metric.go
type RayClusterMetricCollector struct {
// Metrics
rayClusterCreatedTotal: ...
}
func NewRayClusterMetricCollector(...) ... {
}main.go: Initialize the collectors if this feature is enabled. Pass the collectors into controllers.
Example:
| } else if len(headPods.Items) == 0 { | ||
| // Create head Pod if it does not exist. | ||
| logger.Info("reconcilePods: Found 0 head Pods; creating a head Pod for the RayCluster.") | ||
|
|
There was a problem hiding this comment.
It's still possible to be counted multiple times. KubeRay is possible to delete the head Pod if the restart policy is Never.
win5923
force-pushed
the
metrics/ray_clusters_created_total
branch
from
151382b to
34ba8b5
Compare
Signed-off-by: win5923 <ken89@kimo.com>
Signed-off-by: win5923 <ken89@kimo.com>
Signed-off-by: win5923 <ken89@kimo.com>
win5923
force-pushed
the
metrics/ray_clusters_created_total
branch
from
34ba8b5 to
0a3d445
Compare
3 participants
Why are these changes needed?
Add
ray_clusters_created_totalmetric to track the total number of RayClusters created.created_by_ray_job: true/false-
created_by_ray_service: true/false-
namespaceManual test:
Related issue number
Closes #3175
Checks