Problems began to arise as early as 2008, however, as we began to push the boundaries of UFS2 - both in terms of filesystem size and number of inodes (files). Not only did we start to see prohibitively long fsck times but we started to run out of memory address space that fsck could track inodes with.

At some point in 2011, while working with Kirk McKusick, the author of UFS, I wondered out loud what the real, long-term solution was to the problems we were having as we pushed UFS2 well beyond its design … and he simply said “use ZFS”.

So we embarked on a two year testing and development project to stress test and deploy ZFS which culminated in our first production ZFS systems in 2013. It took another four years before we migrated the final UFS2 storage array to ZFS.