After recent refactor/optimization patches on brick's lvm utility class, it is now impossible to create a lvm_type="thin" pool.
How it should work:
Instanciate the LVM object with parameters creat=True and lvm_type="thin"
-> Creates the Volume group
-> Checks if the pool exists by calling an internal method to call the lvm utility "lvs"
-> See that the pool does not exist
-> Create the pool
-> Activate the pool
-> Everyone is happy
What changed:
The internal methods changed, believeing the patch, in order to reduce the number of actuall calls to the lvm utility CLI tools, with the patchset: dcda67053df5dc0240c743537dc1b9c4a3231b61
What does not work anymore:
To check if a pool exists, the code now calls the lvs utility with the volume_group_name/thin_pool path. This has for effect that lvm cannot find the volume, and the processutil code thus raises an exception (lvs returns a code != 0).
Thus, the constructor of the LVM class cannot check if the pool exists or not, failing any volume creation relying on thin-provisionning.
Actually the order of actions is now:
-> Creates the Volume group
-> Checks if the pool exists by calling lvs on the pool
-> lvs cannot find the LV
-> processutil raises an exception
-> The exception isn't caught by anything (and even if it did, it would not be a proper way to detect absence of thin-pool
-> constructor fails.
Fix proposed to branch: master
Review: https:/