-
Notifications
You must be signed in to change notification settings - Fork 3.8k
monitor: improve handling of write failure #1927
Copy link
Copy link
Closed
Labels
area/monitorImpacts monitoring, access logging, flow logging, visibility of datapath traffic.Impacts monitoring, access logging, flow logging, visibility of datapath traffic.kind/enhancementThis would improve or streamline existing functionality.This would improve or streamline existing functionality.priority/mediumThis is considered important, but not urgent.This is considered important, but not urgent.staleThe stale bot thinks this issue is old. Add "pinned" label to prevent this from becoming stale.The stale bot thinks this issue is old. Add "pinned" label to prevent this from becoming stale.
Description
Metadata
Metadata
Assignees
Labels
area/monitorImpacts monitoring, access logging, flow logging, visibility of datapath traffic.Impacts monitoring, access logging, flow logging, visibility of datapath traffic.kind/enhancementThis would improve or streamline existing functionality.This would improve or streamline existing functionality.priority/mediumThis is considered important, but not urgent.This is considered important, but not urgent.staleThe stale bot thinks this issue is old. Add "pinned" label to prevent this from becoming stale.The stale bot thinks this issue is old. Add "pinned" label to prevent this from becoming stale.
So with the current implementation, we are first removing clients when metadata write fails. This is too simplistic and makes the number of listeners count wrong. We should detect this earlier and remove those connections.