block-device doesn't accept whitelist

[Adam Collard]

Setting block-device to a whitelist of many possible block device names leads to a traceback in the charm.

2014-04-27 11:09:27 INFO install Traceback (most recent call last):
2014-04-27 11:09:27 INFO install File "/var/lib/juju/agents/unit-swift-storage-0/charm/hooks/install", line 91, in <module>
2014-04-27 11:09:27 INFO install main()
2014-04-27 11:09:27 INFO install File "/var/lib/juju/agents/unit-swift-storage-0/charm/hooks/install", line 85, in main
2014-04-27 11:09:27 INFO install hooks.execute(sys.argv)
2014-04-27 11:09:27 INFO install File "/var/lib/juju/agents/unit-swift-storage-0/charm/hooks/charmhelpers/core/hookenv.py", line 381, in execute
2014-04-27 11:09:27 INFO install self._hooks[hook_name]()
2014-04-27 11:09:27 INFO install File "/var/lib/juju/agents/unit-swift-storage-0/charm/hooks/install", line 45, in install
2014-04-27 11:09:27 INFO install setup_storage()
2014-04-27 11:09:27 INFO install File "/var/lib/juju/agents/unit-swift-storage-0/charm/hooks/swift_storage_utils.py", line 176, in setup_storage
2014-04-27 11:09:27 INFO install for dev in determine_block_devices():
2014-04-27 11:09:27 INFO install File "/var/lib/juju/agents/unit-swift-storage-0/charm/hooks/swift_storage_utils.py", line 167, in determine_block_devices
2014-04-27 11:09:27 INFO install return [ensure_block_device(bd) for bd in bdevs]
2014-04-27 11:09:27 INFO install File "/var/lib/juju/agents/unit-swift-storage-0/charm/hooks/misc_utils.py", line 57, in ensure_block_device
2014-04-27 11:09:27 INFO install if not is_block_device(bdev):
2014-04-27 11:09:27 INFO install File "/var/lib/juju/agents/unit-swift-storage-0/charm/hooks/charmhelpers/contrib/storage/linux/utils.py", line 17, in is_block_device
2014-04-27 11:09:27 INFO install return S_ISBLK(stat(path).st_mode)
2014-04-27 11:09:27 INFO install OSError: [Errno 2] No such file or directory: '/dev/vdb'

I would expect either is_block_device() to return False if the path doesn't exist, or the call is guarded around a exists() check.

[James Page (james-page)]

We implemented such a feature into the /next branch on rev 30; I'll see if that's SRU'able.

[James Page (james-page)]

My concern with this feature is that this can create a situation where the swift rings are not balanced, and as a result the swift deployment cannot complete; zones in swift rings must have currently have equal numbers of block devices; if you have a zone with less block devices, its not possible to balance the ring.