You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
@@ -74,7 +73,7 @@ The KEP aims at fixing few issues in the current way Events are structured and i
74
73
75
74
Most sections and design details are copied from the original design doc: [Make Kubernetes Events Useful and Safe](https://github.com/kubernetes/community/blob/master/contributors/design-proposals/instrumentation/events-redesign.md).
76
75
77
-
The new Event API has been promoted from v1beta1 to v1 in 1.19 and currently only scheduler was migrated to use it. This KEP proposes to graduate the new Event API by migrating all the remaining Kubernetes components to use the new Event API.
76
+
The new Event API has been promoted from v1beta1 to v1 in 1.19 and currently only scheduler was migrated to use it.
78
77
79
78
## Motivation
80
79
@@ -370,7 +369,7 @@ Scalability and Performance:
370
369
371
370
### Graduation Criteria
372
371
373
-
The new Event API is in v1 right now and scheduler has been migrated to use new API. The plan is to do a load test and migrate all the remaining components.
372
+
The new Event API is in v1 right now and scheduler has been migrated to use new API. The plan is to do a load test before migrating all the remaining components.
374
373
375
374
#### Deprecated Fields
376
375
@@ -383,27 +382,11 @@ This section lists the deprecated Event API fields that should be removed before
383
382
384
383
- Remove deprecated fields listed above
385
384
- Gather data from performance and scalability tests
386
-
- Remove all in-tree use of the core/v1.Event API in favour of events.k8s.io/v1
387
385
388
386
#### Load Test
389
387
390
388
The idea is to use [ClusterLoader](https://github.com/kubernetes/perf-tests/tree/master/clusterloader2) testing framework to do a load test to make sure events generated can be handled by etcd and there's no big performance reduction.
391
389
392
-
#### Components Migration
393
-
394
-
The list of components that need to be migrated is shown as follows:
395
-
396
-
- kubelet
397
-
- cloud-controller-manager
398
-
- kube-controller-manager
399
-
- leader election
400
-
- node problem detector
401
-
- gce ingress controller
402
-
- event exporter
403
-
404
-
(Note: there are more out-of-tree components that need to be migrated)
405
-
406
-
407
390
## Considerations
408
391
409
392
### Performance Impact
@@ -745,4 +728,4 @@ We considered making easier for automated systems to use Events by enforcing "ac
745
728
746
729
### Pivoting Towards Making Events More Helpful for Cluster Operator During Debugging
747
730
748
-
We considered exposing more data that cluster operator would need to use Events for debugging, e.g. making ReportingController more central to the semantics of Event and adding some way to easily grep though the logs of appropriate component when looking for context of a given Event. This idea was dropped because Events are supposed to give application developer who's running his application on the cluster a rough understanding what was happening with his app.
731
+
We considered exposing more data that cluster operator would need to use Events for debugging, e.g. making ReportingController more central to the semantics of Event and adding some way to easily grep though the logs of appropriate component when looking for context of a given Event. This idea was dropped because Events are supposed to give application developer who's running his application on the cluster a rough understanding what was happening with his app.
0 commit comments