Browse Source

Add to rally quotas to handle worst case quota race

Quoting the quota devref:
"""
For a reservation to be successful, the total amount of resources requested,
plus the total amount of resources reserved, plus the total amount of resources
already stored in the database should not exceed the project's quota limit.
"""

This means that in the absolute worst case scenario with 20 concurrent
workers, 19 could have made reservations, committed resources, but not
yet cleared their reservation. Because of the outstanding reservation
and the resources created by the 19 workers, they will all be
double-counted until their reservation is cleared (or it expires).

This adjusts the rally scenarios to handle the double-count for
concurrency.

Related-Bug: #1623390
Change-Id: I4808a92e7e6067aeeb62fc3b3d7f7ac71b179c44
changes/39/371739/4
Kevin Benton 6 years ago
parent
commit
933af3fdac
  1. 10
      rally-jobs/neutron-neutron.yaml

10
rally-jobs/neutron-neutron.yaml

@ -11,7 +11,11 @@
users_per_tenant: 1
quotas:
neutron:
network: 100
# worst case is other 19 writers have created
# resources, but quota reservation hasn't cleared
# yet on any of them. This value could be 100
# without concurrency. see bug/1623390
network: 119
sla:
max_avg_duration_per_atomic:
neutron.list_networks: 15 # reduce as perf is fixed
@ -83,7 +87,9 @@
network: -1
subnet: -1
router: -1
port: 4040 # (ports per network + 1 dhcp) * times
# ((ports per net + 1 dhcp) * times) + (concurrency-1)
# see bug/1623390 for concurrency explanation
port: 4059
sla:
max_avg_duration_per_atomic:
neutron.list_ports: 15 # reduce as perf is fixed

Loading…
Cancel
Save