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
{{ message }}
This repository has been archived by the owner on Jun 20, 2024. It is now read-only.
When there is change in the topology due to addition/deletion of the nodes or transient failures of established connections to the peers, all ODP flows are deleted by InvalidateRoutes
As the topology updates are broadcast to the peers it result in routes calculation multiple times, hence performing InvalidateRoutes
As data point one of the weave-net pod logs of 40 nodes cluster shared in #3641. When a node got disconnected it resulted in performing execess of 100 InvalidateRoutes
When there is change in the topology due to addition/deletion of the nodes or transient failures of established connections to the peers, all ODP flows are deleted by InvalidateRoutes
https://github.com/weaveworks/weave/blob/v2.5.2/router/network_router.go#L70
https://github.com/weaveworks/mesh/blob/v0.2/routes.go#L216
As the topology updates are broadcast to the peers it result in routes calculation multiple times, hence performing InvalidateRoutes
As data point one of the weave-net pod logs of 40 nodes cluster shared in #3641. When a node got disconnected it resulted in performing execess of 100
InvalidateRoutes
weave-net-vjrvd
This invalidating routes would occur across the cluster.
How does this impact? While flows are eventually recreated on the nodes, it leaves window of time where traffic is largely broadcasted.
This feature request is to implement a InvalidateRoutes() in which flows are deleted only if they are impacted by topology change.
The text was updated successfully, but these errors were encountered: