set CINDER_PERIODIC_INTERVAL in devstack to address volume races

After reordering some tests in tempest we exposed a new race. By
default cinder reports free / used space on the cinder periodic
interval (default 60 seconds). In a constrained environment where
we are building and destroy volumes quickly (with secure delete
off) we will get fails out of cinder where it can't build a
volume because it thinks it's out of space (it's not).

The current work around is to decrease the periodic task interval,
the cinder team is working on getting this content updated more
regularly so we wouldn't need this work around in the future.

This race is currently blocking the merge of the tempest.

Needed for bp:tempest-repo-restructure

Change-Id: Ibf49ff6467c871e84fcb716b4762b679b2fb405f
This commit is contained in:
Sean Dague
2013-05-17 06:32:50 -04:00
parent 051b48df41
commit f0161ea7cc

View File

@@ -93,6 +93,7 @@ VIRT_DRIVER=$DEVSTACK_GATE_VIRT_DRIVER
SWIFT_REPLICAS=1
LOG_COLOR=False
PIP_USE_MIRRORS=False
CINDER_PERIODIC_INTERVAL=10
export OS_NO_CACHE=True
EOF