Make SchedulingGroupManager::Leave more efficient
Bug #1461322 reported by
Nischal Sheth
This bug affects 1 person
Affects | Status | Importance | Assigned to | Milestone | ||
---|---|---|---|---|---|---|
Juniper Openstack | Status tracked in Trunk | |||||
R2.20 |
Fix Committed
|
High
|
Nischal Sheth | |||
Trunk |
Fix Committed
|
High
|
Nischal Sheth |
Bug Description
Scaling tests for OVSDB have identified SchedulingGroup
as a heavy CPU user. The tests have a small number of agents that Join
a very large number of VNs. There are 4 TOR-Agents which subscribe to
2K VNs each and 1 TSN-Agent which subscribes to 8K VNs. System takes
23+ minutes to settle down when all these agents are killed. Profiling
data shows that bulk of time is spent in SchedulingGroup
The numbers will be much worse when the number [TOR|TSN]-Agents and/or
VNs goes up.
This bug tracks improvements to handle this scenario more efficiently.
summary: |
- Make SchedulingGroup::Leave more efficient + Make SchedulingGroupManager::Leave more efficient |
description: | updated |
description: | updated |
description: | updated |
To post a comment you must log in.
Review in progress for https:/ /review. opencontrail. org/11209
Submitter: Nischal Sheth (<email address hidden>)