Manually clear backup data for clusters

You can manually clear backup data as needed. OceanBase supports manual backup data clearing only at the cluster level. We recommend that you choose automatic clearing of the backup data in routine use.

You can execute the ALTER SYSTEM statement to clear backup data with the specified backup_set_id, backup_piece_id, and round_id, or clear expired data.

Considerations

• Log archive data clearing depends on data backup. Therefore, before you clear the log archive data, make sure that a data backup file exists. If no data backup file exists, the log archive data cannot be cleared.

• When you clear backup data from the Network File System (NFS), the system directly clears backup files that meet the requirements. When you clear backup data from the Object Storage Service (OSS), the backup file clearing mode is specified by the delete_mode parameter in backup_dest. For more information, see Preparations for backup.

Clear specified backup data

OceanBase Database allows you to clear backup data with the specified backup_set_id, backup_piece_id, and backup_round_id.

1. Log on to the sys tenant as the root user.

2. Query the CDB_OB_BACKUP_SET_FILES, CDB_OB_BACKUP_ARCHIVELOG_SUMMARY, and CDB_OB_BACKUP_PIECE_FILES views to find the backup_set_id, round_id, and backup_piece_id corresponding to the backup data to be cleared.

   • Obtain the backup_set_id of the backup data

     Sample statement for querying the CDB_OB_BACKUP_SET_FILES view:

     obclient> SELECT * FROM oceanbase.CDB_OB_BACKUP_SET_FILES WHERE tenant_id=1 AND copy_id=0 AND file_status != 'DELETED';
     +-------------+-----------+--------+---------+-------------+-----------------+---------+-------------+----------------------------+----------------------------+------------------+------+------------+------------+--------------+-------------------+-------------------+----------------------+---------------------------+--------------------+
     | INCARNATION | TENANT_ID | BS_KEY | COPY_ID | BACKUP_TYPE | ENCRYPTION_MODE | STATUS  | FILE_STATUS | START_TIME                 | COMPLETION_TIME            | ELAPSED_SECONDES | KEEP | KEEP_UNTIL | COMPRESSED | OUTPUT_BYTES | OUTPUT_RATE_BYTES | COMPRESSION_RATIO | OUTPUT_BYTES_DISPLAY | OUTPUT_RATE_BYTES_DISPLAY | TIME_TAKEN_DISPLAY |
     +-------------+-----------+--------+---------+-------------+-----------------+---------+-------------+----------------------------+----------------------------+------------------+------+------------+------------+--------------+-------------------+-------------------+----------------------+---------------------------+--------------------+
     |           1 |         1 |      3 |       0 | D           | NONE            | SUCCESS | AVAILABLE   | 2022-10-28 11:19:56.883788 | 2022-10-28 11:19:59.629990 |                3 | NO   |            | NO         |      3376538 |      1229530.0928 |              0.02 | 3.22MB               | 1.17MB/S                  | 00:00:02.746202    |
     +-------------+-----------+--------+---------+-------------+-----------------+---------+-------------+----------------------------+----------------------------+------------------+------+------------+------------+--------------+-------------------+-------------------+----------------------+---------------------------+--------------------+
     1 row in set
     

     Fields in query results are as follows:

     • BS_KEY: indicates the backup set ID of the data backup, which is also backup_set_id.

     • COPY_ID: The value of the COPY_ID field is 0 for a data backup task.

     For more information about the fields in the CDB_OB_BACKUP_SET_FILES view, see oceanbase.CDB_OB_BACKUP_SET_FILES.

   • Obtain the round_id of the log backup

     obclient> SELECT * FROM oceanbase.CDB_OB_BACKUP_ARCHIVELOG_SUMMARY WHERE tenant_id=1;
     +-------------+-------------------+-----------+--------+----------------+-----------------+----------------------------+----------------------------+-------------+--------------+-------------------+---------------------+----------------------+
     | INCARNATION | LOG_ARCHIVE_ROUND | TENANT_ID | STATUS | START_PIECE_ID | BACKUP_PIECE_ID | MIN_FIRST_TIME             | MAX_NEXT_TIME              | INPUT_BYTES | OUTPUT_BYTES | COMPRESSION_RATIO | INPUT_BYTES_DISPLAY | OUTPUT_BYTES_DISPLAY |
     +-------------+-------------------+-----------+--------+----------------+-----------------+----------------------------+----------------------------+-------------+--------------+-------------------+---------------------+----------------------+
     |           1 |                 1 |         1 | STOP   |              0 |               0 | 2022-10-24 18:02:10.951125 | 2022-10-24 19:43:06.867056 |           0 |            0 |              NULL | 0.00MB              | 0.00MB               |
     |           1 |                 2 |         1 | DOING  |              0 |               0 | 2022-10-24 19:43:27.652015 | 2022-10-24 19:46:39.949493 |           0 |            0 |              NULL | 0.00MB              | 0.00MB               |
     +-------------+-------------------+-----------+--------+----------------+-----------------+----------------------------+----------------------------+-------------+--------------+-------------------+---------------------+----------------------+
     2 rows in set
     

     The value in the LOG_ARCHIVE_ROUND column is the round_id that you require. Only a round in the STOP state can be cleared.

     For more information about the fields in the CDB_OB_BACKUP__ARCHIVELOG_SUMMARY view, see oceanbase.CDB_OB_BACKUP_ARCHIVELOG_SUMMARY.

   • Obtain the corresponding backup_piece_id based on the round_id of the log backup

     Sample statement for querying the oceanbase.CDB_OB_BACKUP_PIECE_FILES view:

     obclient> SELECT * FROM oceanbase.CDB_OB_BACKUP_PIECE_FILES WHERE tenant_id=1 AND round_id=1 and copy_id=0 AND status!='ACTIVE' AND file_status!='DELETED';
     +-------------+-----------+----------+-----------------+---------+-------------+----------------------------+----------------------------+----------------------------+--------+-------------+------------+----------------+
     | INCARNATION | TENANT_ID | ROUND_ID | BACKUP_PIECE_ID | COPY_ID | CREATE_DATE | START_TS                   | CHECKPOINT_TS              | MAX_TS                     | STATUS | FILE_STATUS | COMPATIBLE | START_PIECE_ID |
     +-------------+-----------+----------+-----------------+---------+-------------+----------------------------+----------------------------+----------------------------+--------+-------------+------------+----------------+
     |           1 |         1 |        1 |               0 |       0 |    20221028 | 2022-10-28 10:58:38.325595 | 2022-10-28 14:50:29.834598 | 2022-10-28 14:50:39.280049 | FROZEN | AVAILABLE   |          1 |              0 |
     +-------------+-----------+----------+-----------------+---------+-------------+----------------------------+----------------------------+----------------------------+--------+-------------+------------+----------------+
     1 row in set
     

     Fields in query results are as follows:

     • ROUND_ID: indicates the round_id of the log backup, if the round-based backup is enabled.

     • BACKUP_PIECE_ID: indicates the piece ID of the log backup. If log splitting is enabled, backup_piece_id > 0. If log splitting is disabled, backup_piece_id = 0.

     • COPY_ID: The value of the COPY_ID field is 0 for a log backup task. This view contains information about the backup and secondary backup tasks.

     For more information about the fields in the CDB_OB_BACKUP_PIECE_FILES view, see oceanbase.CDB_OB_BACKUP_PIECE_FILES.

3. Execute the statement as needed to clear the backup data.

   • Clear a backup set with the specified backup set ID (backup_set_id).

     obclient> ALTER SYSTEM DELETE BACKUPSET backup_set_id;
     

     For example, you can execute the following statement to clear the backup set whose backup set ID is 1:

     obclient> ALTER SYSTEM DELETE BACKUPSET 1;
     

     When you execute the DELETE BACKUPSET statement to clear backup sets, the following limitations apply:

     • When you clear backup sets of the specified backup set IDs, if the destinations of the backup sets are the same as the destination specified for backup_dest, at least one complete and valid backup copy is retained. If their destinations are different from the specified destination, all data is cleared.

     • When you clear backup sets based on the specified backup set IDs, if the destinations of the backup sets are the same as the destination specified for backup_dest, you must clear the backup sets based on the order of their backup set IDs. For example, when you clear three backup sets whose backup set IDs are 1, 2, and 3, you must clear the backup sets in the order of 1, 2, and 3. If their destinations are different from the specified destination, you can clear the backup sets without following the preceding order.

     • The operation fails if a specified backup set ID does not exist or a backup task with a specified backup set ID is still in progress.

   • Clear all data in a log backup round with the specified round_id.

     obclient> ALTER SYSTEM DELETE BACKUPROUND round_id;
     

     The limitations on the DELETE BACKUPPIECE command also apply to the DELETE BACKUPROUND command:

     • You can only delete backup rounds in the STOP state.

     • When you clear backup data of the specified round IDs, if the destinations of the backup rounds are the same as the destination specified for backup_dest and the existing data does not fully cover the backup rounds, you cannot clear backup data of the corresponding round IDs. If their destinations are different from the specified destination, you can clear the backup data of the specified round IDs regardless of whether the existing backup data fully covers the backup rounds.

     • When you clear backup data of the specified round IDs, if the destinations of the backup rounds are the same as the destination specified for backup_dest, you must clear the backup data based on the order of the backup round IDs to avoid log discontinuity. For example, when you clear backup data whose round IDs are 1, 2, and 3, you must clear the backup data in the order of 1, 2, and 3. If their destinations are different from the specified destination, you can clear the backup data of the specified round IDs without following the preceding order.

   • Clear archive data of the specified backup piece IDs.

     obclient> ALTER SYSTEM DELETE BACKUPPIECE backup_piece_id;
     

     When you run the DELETE BACKUPPIECE command to clear backup pieces, the following limitations apply:

     • You can run this command to clear specified pieces only when log splitting is enabled for log backup.

       For more information about how to enable log splitting for log backup, see Enable backup based on log splitting.

     • You can clear only backup pieces that are in the Frozen or Inactive state.

     • When you clear backup pieces of the specified piece IDs, if the destinations of the backup pieces are the same as the destination specified for backup_dest and the existing backup data does not fully cover the backup pieces, you cannot clear the backup pieces of the corresponding piece IDs. If their destinations are different from the specified destination, you can clear the backup pieces of the specified piece IDs regardless of whether the existing backup data fully covers the backup pieces.

     • When you clear backup pieces of the specified piece IDs, if the destinations of the backup pieces are the same as the destination specified for backup_dest, you must clear the backup pieces based on the order of the backup piece IDs to avoid log discontinuity. For example, when you clear backup pieces whose piece IDs are 1, 2, and 3, you must clear the backup pieces in the order of 1, 2, and 3. If their destinations are different from the specified destination, you can clear the backup pieces of the specified piece IDs without following the preceding order.

Clear expired backup data

Preparations

Before you manually clear expired backup data, configure the clearing strategy for the backup data. That is, set the recovery_window parameter in backup_dest_option or set the backup_recovery_window parameter.

The recovery_window parameter specifies the time window for restoring backup data. The system determines whether backup data has expired based on the value of this parameter.

All backup data earlier than the latest backup data that meets the following condition is considered expired: current time - backup point in time >= recovery_window.

For example, the setting recovery_window='7d' indicates that the data backed up within the last 7 days must be restorable, and the backup data beyond this restore window is considered expired.

The backup_recovery_window parameter in this statement is equivalent to the recovery_window parameter in backup_dest_option. For more information about the recovery_window parameter, see Automatically clear backup data for clusters.

For more information about the backup_dest_option parameter, see backup_dest_option.

Manually clear expired backup data

1. Log on to the sys tenant as the root user.

2. Execute the following statement to clear expired backup data:

   obclient> ALTER SYSTEM DELETE OBSOLETE BACKUP;
   

   Note the following limitations for using the DELETE OBSOLETE command:

   • If no or only one valid backup set exists, you cannot clear the backup pieces of the log archive data.

   • The DELETE OBSOLETE command only clears expired backup data whose storage path is the same as the destination specified for backup_dest. For more information about expired backup data whose destination is changed, see Clear specified backup data in this topic.

   • The DELETE OBSOLETE command can clear secondary backup data corresponding to the backup_dest parameter.

   • The operation fails if the number of existing backup copies is less than the value specified for the backup_copies parameter for backup_dest_option.

What to do next

You can also configure the automatic clearing of backup data. For more information, see Automatically clear backup data for clusters.

OceanBase Database allows you to stop backup data clearing. For more information, see Stop backup data clearing for clusters.