So far my experience with cephfs has been really great. Comparing it to other systems I use like gluster, Cephfs is really coming out on top for me. And the flexibility of cephfs is one I wanted to hit on today.
Imagine if you will the topology of a standard cephfs data pool named cephfs_pool.
Inside there are 5 directories created for all sorts of varying purposes. In the production directory, performance is of the utmost importance to us, as opposed to backups where we really just need a dumping ground for data.
Instead of everyone using the same pool (cephfs_pool), we can actually map these different directories to other pools where our specific needs can be met.
Lets pretend these are the pools we are working with.
/production to the
high_performace_pool. We’ll do this by using the
First mount the root cephfs directory
Then go into that new mount point and change each subdirectory to the pool you want its data to reside on.
Now to verify it worked.
And thats it! Your data will now reside on what ever pool you have changed this to.
Note - Files inherit the layout of their parent directory at creation time. However, subsequent changes to the parent directory’s layout do not affect children.