-
Story
-
Resolution: Done
-
Undefined
-
None
-
None
-
None
-
Future Sustainability
-
False
-
-
False
-
None
-
None
-
None
-
NetObserv - Sprint 263, NetObserv - Sprint 264, NetObserv - Sprint 265
Follow-up on using a global hmap for TC instead of per-CPU, it becomes possible to move the dedup logic from user space to kernel space. Roughly adopting the same logic: first interface seeing a flow prevails for adding counters - other interfaces don't update counters.
MACs will be removed from flow_id as part of this.
- depends on
-
NETOBSERV-1995 Move away from per-CPU maps
-
- Closed
-
- is depended on by
-
NETOBSERV-1966 Use packet tracing for deduplication
-
- Closed
-
- links to