Removing a Fast Data Directory or Large Data Directory From a Forest
27 November 2019 05:11 AM
MarkLogic Server offers Fast Data Directories, and Large Data Directories to allow customers to better utilize their available infrastructure. This allows an organization to offload large objects to cheaper storage, or improve performance with SSDs for portions of a forest. These directories are defined at the forest level, usually when the forest was created.
Removing Fast or Large Data Directories
There are two primary methods to remove these directories from a forest.
Rebalancing to a New Forest
This method takes advantage of the rebalancing mechanism in the server to move data from the forest with the Fast/Large Data Directories. New forests can be defined as part of this process, but it is not required. The advantage of this method is that it does not require any downtime. The primary disadvantage is that in can increase the IO, and CPU load on the servers as the data is moved between forests, and can result in data being moved more than once. If needed, these issues can be mitigate by adjusting the rebalancer priority and merge settings.
Backup/Restore to a new forest
This method allows a simple 1 for 1 swap of a forest with a Fast/Large Data Directory to one without these directories. The advantage of this method is that, depending on the size of the forest, it can be completed faster than rebalancing. There are a couple of disadvantages to this method. The first is that the forest being replaced needs to be in read only mode when the backup is taken, until the restore is complete to the new forest. The second is that it does require some downtime when switching between the old and new forests. These issues can be mitigated with some careful planning.
Procedures for Using Rebalance
Procedures for Using Backup/Restore