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
Copy file name to clipboardExpand all lines: keps/sig-cloud-provider/5237-watch-based-route-controller-reconciliation/README.md
+4-4Lines changed: 4 additions & 4 deletions
Original file line number
Diff line number
Diff line change
@@ -204,7 +204,7 @@ As a cluster operator I need to use the API rate limits from my infrastructure p
204
204
205
205
We currently use the `Node.Status.Addresses` and `PodCIDRs` fields to trigger updates in the route reconciliation mechanism. However, relying solely on these fields may be insufficient, potentially causing missed route reconciliations when updates are necessary. This depends on the specific cloud-controller-manager implementations. Using these fields works for the CCM maintained by the authors, but we do not know the details of other providers.
206
206
207
-
This is mitigated by a feature gate, which allows infrastructure providers to test it and provide feedback on the fields.
207
+
This is mitigated by the feature gate`CloudControllerManagerWatchBasedRoutesReconciliation`, which allows infrastructure providers to test it and provide feedback on the fields.
208
208
209
209
#### Relying only on Events
210
210
@@ -219,7 +219,7 @@ This is mitigated by:
219
219
220
220
## Design Details
221
221
222
-
We use a similar concept as already implemented in the node and service controller. Through node informers we register event handlers for the add, delete and update node events, where updates are filtered by certain criteria. To introduce the feature we establish a new feature flag called `CloudControllerManagerWatchBasedRoutesReconciliation`.
222
+
We use a similar concept as already implemented in the node and service controller. Through node informers we register event handlers for the add, delete and update node events, where updates are filtered by certain criteria. To introduce the feature we establish a new feature gate called `CloudControllerManagerWatchBasedRoutesReconciliation`.
223
223
224
224
#### Full reconcile
225
225
@@ -241,8 +241,8 @@ Node updates are quite frequent. To reduce the number of requests sent to infras
241
241
242
242
Two fields are relevant for determining whether a reconcile should occur:
243
243
244
-
1.`Node.Status.Addresses` maps to the `TargetNodeAddresses` field in the `Route` struct. It determines where packets for a give IP range should be sent. Changes to this field must trigger a reconcile.
245
-
2.`Node.Spec.PodCIDRs` contains the IP ranges assigned to the node. These CIDRs are used as the destination in the created routes.. Changes to this field must trigger a reconcile.
244
+
1.`Node.Status.Addresses` maps to the `TargetNodeAddresses` field in the `Route` struct. It determines where packets for a given IP range should be sent. Changes to this field must trigger a reconcile.
245
+
2.`Node.Spec.PodCIDRs` contains the IP ranges assigned to the node. These CIDRs are used as the destination in the created routes. Changes to this field must trigger a reconcile.
0 commit comments