|
| 1 | +from typing import Tuple |
| 2 | + |
| 3 | +from google.api_core import operation as api_core_operation |
| 4 | + |
| 5 | +from google.cloud.bigtable import admin_v2 |
| 6 | +from google.cloud.bigtable.data._cross_sync import CrossSync |
| 7 | +from google.cloud.bigtable.data import mutations, read_rows_query |
| 8 | + |
| 9 | +import time |
| 10 | + |
| 11 | +from .conftest import ( |
| 12 | + INSTANCE_PREFIX, |
| 13 | + BACKUP_PREFIX, |
| 14 | + ROW_PREFIX, |
| 15 | + DEFAULT_CLUSTER_LOCATIONS, |
| 16 | + REPLICATION_CLUSTER_LOCATIONS, |
| 17 | + TEST_TABLE_NAME, |
| 18 | + TEST_BACKUP_TABLE_NAME, |
| 19 | + TEST_COLUMMN_FAMILY_NAME, |
| 20 | + TEST_COLUMN_NAME, |
| 21 | + NUM_ROWS, |
| 22 | + INITIAL_CELL_VALUE, |
| 23 | + NEW_CELL_VALUE, |
| 24 | +) |
| 25 | + |
| 26 | +from datetime import datetime, timedelta |
| 27 | + |
| 28 | +import pytest |
| 29 | + |
| 30 | +# TODO: Finish CrossSync integration after moving to new directory |
| 31 | + |
| 32 | + |
| 33 | +_TABLE_ADMIN_CLIENT = admin_v2.BigtableTableAdminClient() |
| 34 | +_INSTANCE_ADMIN_CLIENT = admin_v2.BigtableInstanceAdminClient() |
| 35 | +_DATA_CLIENT = CrossSync._Sync_Impl.DataClient() |
| 36 | + |
| 37 | + |
| 38 | +@pytest.fixture(scope="function") |
| 39 | +def instances_to_delete(): |
| 40 | + instances_to_delete = [] |
| 41 | + |
| 42 | + yield instances_to_delete |
| 43 | + |
| 44 | + for instance in instances_to_delete: |
| 45 | + _INSTANCE_ADMIN_CLIENT.delete_instance(name=instance.name) |
| 46 | + |
| 47 | + |
| 48 | +@pytest.fixture(scope="function") |
| 49 | +def backups_to_delete(): |
| 50 | + backups_to_delete = [] |
| 51 | + |
| 52 | + yield backups_to_delete |
| 53 | + |
| 54 | + for backup in backups_to_delete: |
| 55 | + _TABLE_ADMIN_CLIENT.delete_backup(name=backup.name) |
| 56 | + |
| 57 | + |
| 58 | +def ctime() -> int: |
| 59 | + return int(time.time()) |
| 60 | + |
| 61 | + |
| 62 | +def create_instance( |
| 63 | + project_id, |
| 64 | + instances_to_delete, |
| 65 | + storage_type=admin_v2.StorageType.HDD, |
| 66 | + cluster_locations=DEFAULT_CLUSTER_LOCATIONS, |
| 67 | +) -> Tuple[admin_v2.Instance, admin_v2.Table]: |
| 68 | + clusters = {} |
| 69 | + |
| 70 | + instance_id = f"{INSTANCE_PREFIX}-{ctime()}" |
| 71 | + |
| 72 | + for idx, location in enumerate(cluster_locations): |
| 73 | + clusters[location] = admin_v2.Cluster( |
| 74 | + name=_INSTANCE_ADMIN_CLIENT.cluster_path(project_id, instance_id, f"{instance_id}-{idx}"), |
| 75 | + location=_INSTANCE_ADMIN_CLIENT.common_location_path(project_id, location), |
| 76 | + default_storage_type=storage_type |
| 77 | + ) |
| 78 | + |
| 79 | + create_instance_request = admin_v2.CreateInstanceRequest( |
| 80 | + parent=_INSTANCE_ADMIN_CLIENT.common_project_path(project_id), |
| 81 | + instance_id=instance_id, |
| 82 | + instance=admin_v2.Instance( |
| 83 | + display_name=instance_id[:30], # truncate to 30 characters because of character limit |
| 84 | + ), |
| 85 | + clusters=clusters |
| 86 | + ) |
| 87 | + operation = _INSTANCE_ADMIN_CLIENT.create_instance(create_instance_request) |
| 88 | + instance = operation.result() |
| 89 | + |
| 90 | + instances_to_delete.append(instance) |
| 91 | + |
| 92 | + create_table_request = admin_v2.CreateTableRequest( |
| 93 | + parent=_INSTANCE_ADMIN_CLIENT.instance_path(project_id, instance_id), |
| 94 | + table_id=TEST_TABLE_NAME, |
| 95 | + table=admin_v2.Table( |
| 96 | + column_families={ |
| 97 | + TEST_COLUMMN_FAMILY_NAME: admin_v2.ColumnFamily(), |
| 98 | + } |
| 99 | + ) |
| 100 | + ) |
| 101 | + |
| 102 | + table = _TABLE_ADMIN_CLIENT.create_table(create_table_request) |
| 103 | + |
| 104 | + # Populate with dummy data |
| 105 | + populate_table(instance, table, INITIAL_CELL_VALUE) |
| 106 | + |
| 107 | + return instance, table |
| 108 | + |
| 109 | + |
| 110 | +def populate_table(instance, table, cell_value): |
| 111 | + data_client_table = _DATA_CLIENT.get_table( |
| 112 | + _TABLE_ADMIN_CLIENT.parse_instance_path(instance.name)["instance"], |
| 113 | + _TABLE_ADMIN_CLIENT.parse_table_path(table.name)["table"], |
| 114 | + ) |
| 115 | + row_mutation_entries = [] |
| 116 | + for i in range(0, NUM_ROWS): |
| 117 | + row_mutation_entries.append( |
| 118 | + mutations.RowMutationEntry( |
| 119 | + row_key=f"{ROW_PREFIX}-{i}", |
| 120 | + mutations=[ |
| 121 | + mutations.SetCell( |
| 122 | + family=TEST_COLUMMN_FAMILY_NAME, |
| 123 | + qualifier=TEST_COLUMN_NAME, |
| 124 | + new_value=cell_value, |
| 125 | + timestamp_micros=-1 |
| 126 | + ) |
| 127 | + ] |
| 128 | + ) |
| 129 | + ) |
| 130 | + |
| 131 | + data_client_table.bulk_mutate_rows(row_mutation_entries) |
| 132 | + |
| 133 | + |
| 134 | +def create_backup(instance, table, backups_to_delete) -> admin_v2.Backup: |
| 135 | + list_clusters_response = _INSTANCE_ADMIN_CLIENT.list_clusters(parent=instance.name) |
| 136 | + cluster_name = list_clusters_response.clusters[0].name |
| 137 | + |
| 138 | + backup_id = f"{BACKUP_PREFIX}-{ctime()}" |
| 139 | + |
| 140 | + operation = _TABLE_ADMIN_CLIENT.create_backup( |
| 141 | + admin_v2.CreateBackupRequest( |
| 142 | + parent=cluster_name, |
| 143 | + backup_id=backup_id, |
| 144 | + backup=admin_v2.Backup( |
| 145 | + name=f"{cluster_name}/backups/{backup_id}", |
| 146 | + source_table=table.name, |
| 147 | + expire_time=datetime.now() + timedelta(hours=7) |
| 148 | + ) |
| 149 | + ) |
| 150 | + ) |
| 151 | + |
| 152 | + backup = operation.result() |
| 153 | + backups_to_delete.append(backup) |
| 154 | + return backup |
| 155 | + |
| 156 | + |
| 157 | +def assert_table_cell_value_equal_to(instance, table, value): |
| 158 | + data_client_table = _DATA_CLIENT.get_table( |
| 159 | + _TABLE_ADMIN_CLIENT.parse_instance_path(instance.name)["instance"], |
| 160 | + _TABLE_ADMIN_CLIENT.parse_table_path(table.name)["table"], |
| 161 | + ) |
| 162 | + |
| 163 | + # Read all the rows; there shouldn't be that many of them |
| 164 | + query = read_rows_query.ReadRowsQuery(limit=NUM_ROWS) |
| 165 | + for row in data_client_table.read_rows_stream(query): |
| 166 | + latest_cell = row[TEST_COLUMMN_FAMILY_NAME, TEST_COLUMN_NAME][0] |
| 167 | + assert latest_cell.value.decode("utf-8") == value |
| 168 | + |
| 169 | + |
| 170 | +@pytest.mark.parametrize( |
| 171 | + "second_instance_storage_type,expect_optimize_operation", |
| 172 | + [ |
| 173 | + (admin_v2.StorageType.HDD, False), |
| 174 | + (admin_v2.StorageType.SSD, True), |
| 175 | + ] |
| 176 | +) |
| 177 | +def test_optimize_restored_table( |
| 178 | + admin_overlay_project_id, |
| 179 | + instances_to_delete, |
| 180 | + backups_to_delete, |
| 181 | + second_instance_storage_type, |
| 182 | + expect_optimize_operation, |
| 183 | +): |
| 184 | + # Create two instances. We backup a table from the first instance to a new table in the |
| 185 | + # second instance. This is to test whether or not different scenarios trigger an |
| 186 | + # optimize_restored_table operation |
| 187 | + instance_with_backup, table_to_backup = create_instance( |
| 188 | + admin_overlay_project_id, |
| 189 | + instances_to_delete, |
| 190 | + admin_v2.StorageType.HDD |
| 191 | + ) |
| 192 | + |
| 193 | + instance_to_restore, _ = create_instance( |
| 194 | + admin_overlay_project_id, |
| 195 | + instances_to_delete, |
| 196 | + second_instance_storage_type, |
| 197 | + ) |
| 198 | + |
| 199 | + backup = create_backup(instance_with_backup, table_to_backup, backups_to_delete) |
| 200 | + |
| 201 | + # Restore to other instance |
| 202 | + restore_operation = _TABLE_ADMIN_CLIENT.restore_table( |
| 203 | + admin_v2.RestoreTableRequest( |
| 204 | + parent=instance_to_restore.name, |
| 205 | + table_id=TEST_BACKUP_TABLE_NAME, |
| 206 | + backup=backup.name, |
| 207 | + ) |
| 208 | + ) |
| 209 | + |
| 210 | + assert isinstance(restore_operation, admin_v2.RestoreTableOperation) |
| 211 | + restored_table = restore_operation.result() |
| 212 | + |
| 213 | + optimize_operation = restore_operation.optimize_restored_table_operation() |
| 214 | + if expect_optimize_operation: |
| 215 | + assert isinstance(optimize_operation, api_core_operation.Operation) |
| 216 | + optimize_operation.result() |
| 217 | + else: |
| 218 | + assert optimize_operation is None |
| 219 | + |
| 220 | + # Test that the new table exists |
| 221 | + assert restored_table.name == f"{instance_to_restore.name}/tables/{TEST_BACKUP_TABLE_NAME}" |
| 222 | + assert_table_cell_value_equal_to(instance_to_restore, restored_table, INITIAL_CELL_VALUE) |
| 223 | + |
| 224 | + |
| 225 | +def test_wait_for_consistency(admin_overlay_project_id, instances_to_delete): |
| 226 | + instance, table = create_instance( |
| 227 | + admin_overlay_project_id, |
| 228 | + instances_to_delete, |
| 229 | + cluster_locations=REPLICATION_CLUSTER_LOCATIONS, |
| 230 | + ) |
| 231 | + |
| 232 | + populate_table(instance, table, NEW_CELL_VALUE) |
| 233 | + wait_for_consistency_request = admin_v2.WaitForConsistencyRequest( |
| 234 | + name=table.name, |
| 235 | + ) |
| 236 | + wait_for_consistency_request.standard_read_remote_writes = admin_v2.StandardReadRemoteWrites() |
| 237 | + |
| 238 | + _TABLE_ADMIN_CLIENT.wait_for_consistency(wait_for_consistency_request) |
| 239 | + assert_table_cell_value_equal_to(instance, table, NEW_CELL_VALUE) |
0 commit comments