At FB, we use containers everywhere. This talk covers how we use the advanced features of Btrfs to deploy and manage container filesystems across our fleet. It will cover btrfs basics, from internal structures, how to we subvolumes, and how we create and manage snapshots, to more advanced features including: how we create btrfs sub-volume diffs for shipping incremental filesystems. In addition I will cover how we utilize this diff mechanism to manage image layering. I will also cover in this talk some gotchas and problems that we've discovered in our usage of btrfs at significant scale.

Attendees will be excited to hear about it as Btrfs has had a lot of interest over the years but there hasn't been very many large scale distributions of btrfs that showcase the value that it can provide.  I will talk about how btrfs at scale can work and how it has worked for the Facebook infrastructure. 

Attendees shouild be familar with btrfs and it's basic operations.  Indepth knowledge of the filesystem internals is not required, but a basic undrestanding of it's feature set would be expected.

My session will describe a method by with container filesystems can be deployed incrementaly across large fleets of servers.  This is not specific just to Facebook infrastructure, the methods I will show can be used with various different container runtimes.