This appears to have been broken, which means we weren't actually testing
any of the non-default cache modes (writethrough, none). Whoops! We were
getting jobs like
task:
- install: null
- ceph:
conf:
client:
rbd cache: false
tasks:
- chef: null
- clock: null
- install: null
- ceph: null
- rbd_fsx:
clients:
- client.0
ops: 5000
Signed-off-by: Sage Weil <sage@inktank.com>
-task:
+tasks:
- install:
- ceph:
conf:
-task:
+tasks:
- install:
- ceph:
conf:
-task:
+tasks:
- install:
- ceph:
conf:
+++ /dev/null
-tasks:
-- install:
-- ceph: