Why does the Zeppelin node have a data disc

[Zarquan]

Need to check why the Zeppelin node has a Cinder volume attached.

The worker nodes use Cinder volumes to store their HDFS data, but the Zeppelin node isn't running the HDFS daemon, so it doesn't need it for that.

If we don't need the extra volume, then we can remove it from the Ansible scripts.

[Zarquan]

I think Zeppelin needs a Cinder volume because it is acting as the Spark master, and needs extra space for the Spark temp data.

[Zarquan]

This isn't an issue as such, the Zeppelin node needs the extra space, but the space should be on a local disc not a Cinder volume.
Working on this at the moment.

[Zarquan]

Because Zeppelin interpreter is acting as the Spark master, the Spark temp data gets stored on the Zeppelin node.

[Zarquan]

I think the discussion is done - the Spark master needs a data disc, and in our current configuration the Zeppelin node is our Spark master.

[Zarquan]

If we could convert the discussion back to an issue I'd close it.