Table and Cluster Admin APIs (fully working and tested)

This is an alias for:
https://github.com/dhermes/gcloud-python-bigtable/releases/tag/table-api-complete

Table and Cluster Admin APIs (fully working and tested)

We have moved from a connection based approach to a Client hierarchy.

See http://gcloud-python-bigtable.readthedocs.org/en/table-api-complete/ for docs at this release.

Cluster Admin API (fully working and tested)

The Cluster Admin API has been fully implemented in this release.

To use the API, the Cluster class defines a high-level interface:

from gcloud_bigtable.cluster import Cluster

cluster = Cluster(project_id, zone, cluster_id)

Create a new Cluster

After creating the cluster object, make a CreateCluster API request:

display_name = 'My very own cluster'
cluster.create(display_name)

If you would like more than the minimum number of nodes (3) in your cluster:

serve_nodes = 10
cluster.create(display_name, serve_nodes=serve_nodes)

NOTE: Currently serve_nodes will default to 3 if not passed to create and display_name is mandatory in create since it is mandatory in CreateCluster. However, we could easily set

display_name = cluster.cluster_id

if cluster.create() were called with no arguments.

Get metadata for an existing Cluster

After creating the cluster object, make a GetCluster API request:

cluster.reload()

This will load serve_nodes and display_name for the existing cluster in addition to the cluster_id, zone and project_id already set on the Cluster object.

Update an existing Cluster

After creating the cluster object, make an UpdateCluster API request:

cluster.update(display_name=display_name, serve_nodes=serve_nodes)

Both display_name and serve_nodes are optional arguments. If neither are passed in, the method will do nothing.

NOTE: This means that if you change them via:

cluster.display_name = 'Brand new display name'
cluster.update()

then the update won't actually occur. (Though we are open to changing this behavior. The current behavior is in place to avoid an HTTP/2 request if one is not necessary. This behavior would not be possible if display_name and serve_nodes were read-only properties.)

Delete an existing Cluster

Make a DeleteCluster API request:

cluster.delete()

Low-level Methods

The ListClusters, UndeleteCluster, and ListZones methods
have been implemented on the ClusterConnection class, but not on
the Cluster convenience class.

For now, you can access a cluster connection as a protected attribute of a Cluster:

cluster_connection = cluster._cluster_conn

# Returns a
# (gcloud_bigtable._generated.bigtable_cluster_service_messages_pb2.
#  ListClustersResponse)
list_of_clusters = cluster_connection.list_clusters(project_id)

# Returns a 
# gcloud_bigtable._generated.operations_pb2.Operation
long_running_operation = cluster_connection.undelete_cluster(
    project_id, zone, cluster_id)

If you don't know which zone you want the cluster in, create a
Cluster without one and then make a ListZones request
to choose one:

cluster = Cluster(project_id, None, cluster_id)
cluster_connection = cluster._cluster_conn

# Returns a 
# (gcloud_bigtable._generated.bigtable_cluster_service_messages_pb2.
#  ListZonesResponse)
list_zones_response = cluster_connection.list_zones(
    cluster.cluster_id)
zone_choices = [zone.display_name 
                for zone in list_zones_response.zones]

# Found the best choice! The 4th zone.
cluster.zone = zone_choices[3]

Low-level Responses

These are low-level because list_of_clusters will be a
ListClustersResponse, long_running_operation will be a
long-running Operation and list_zones_response will be a
ListZonesResponse