drop() hangs

drop() hangs in the Gremlin console.

Dropping a graph or a portion of a graph such as some vertices can hang based on errors in logged/unlogged or causes error depending on logged/unlogged Cassandra batches. A method to resolve the problem is to drop or truncate the underlying Cassandra tables storing graph data.

The data for a graph is stored in <graph_name>.<vertex_label>_p and <graph_name>.<vertex_label>_e. For example, recipe data stored in a graph food will be food.recipe_p and food.recipe_e.

In some cases, additional steps must be taken to delete a graph. A graph consists of three Cassandra keyspaces: <graph_name>, <graph_name>_system, and <graph_name>_pvt. All three keyspaces must be deleted, with cqlsh if necessary, to completely delete the graph.

If a graph hangs during provisioning, use the following cqlsh commands:
cqlsh> delete from dse_system.shared_data where dataspace = 'Cluster' 
 and valid_until = 13814000-1dd2-11b2-0000-000000000000 
 and namespace = 'system' and name = '<graph_name>';
cqlsh> update dse_system.shared_data set last_updated = now() where dataspace = 'Cluster';
DANGER
Shared_data is not normally manually updated. However, this procedure can be used in the case of a node failure during a graph provisioning operation.