Details
-
New Feature
-
Status: Open
-
Major
-
Resolution: Unresolved
-
None
-
None
Description
We have identified several issues with deploying hadoop + spark bundle and/or hbase + hadoop bundle due to the fact that Juju uses a single root ephemeral disk for installation.
It would be desirable for the Bigdata components to support deploying with additional Juju storage (https://docs.jujucharms.com/storage) by implementing the storage bindings, some issues that should be solved by the additional storage are
- Ability to split application data from the operating system install.
- Backup / Restore ability
- For providers such as OpenStack as there is no Cinder volume, we cannot dynamically expand the volume, so dynamic storage will provide this ability.
- We cannot specify the size of the database at provision time unless we create specific machine flavors for every permutation of disk (eg database) size and cpu/ram
Thank you for your input on this bug.