Red Hat Bugzilla – Bug 1266527
no documentation of partitioning when master+etcd is also an active node
Last modified: 2017-12-30 21:44:54 EST
There is currently no documentation of putting etcd's data on a separate partition.
In certain cases when a user has a Master with etcd that also is acting as an active/schedulable node, docker images, logging, and other activities can fill up the partition that etcd is on.
This can result (in 3.0.2 and earlier) in un-recoverable corruption of the etcd data.
Even when etcd is fixed to safely handle out-of-disk, it's probably just good practice for etcd data to be on a dedicated partition.