Users who wish to have device fault tolerance for their thinpool logical volumes should use the RAID segment types for this purpose (e.g. "raid1"). This is especially encouraged for thinpool metadata. The 'lvconvert' command can be used for this purpose. Here is an example of converting the metadata portion of a thinpool named "my_thinpool" to the "raid1" segment type:
~> lvconvert --type raid1 -m 1 my_vg/my_thinpool_tmeta
The 'raid1' segment type is the new implementation of mirroring in LVM. The legacy mirror segment type is called 'mirror'. Conversions that result in thinpools layered on logical volumes of 'mirror' segment type are no longer allowed. That is, it is no longer possible to create thinpools on top of logical volumes of 'mirror' segment type. This is due to the possibility of I/O hangs and a failure to complete repairs during failure events. Users can still gain the desired fault tolerance by using the 'raid1' segment type which does not suffer from the same limitations.
Users who have already created thinpools with data or metadata areas of 'mirror' segment type will still be able to activate those logical volumes. However, they should convert them to the 'raid1' segment type as soon as possible. This can be quickly accomplished via the 'lvconvert' command. For example, the following command would convert the data portion of a thinpool named 'my_thinpool' in the volume group 'my_vg' from the 'mirror' segment type to the newer 'raid1' segment type:
~> lvconvert --type raid1 my_vg/my_thinpool_tdata