Point-in-time recovery (PITR) allows you to restore the state of your cluster's data and some types of metadata from a specific point in time. This can be relative, such as "three hours ago", or an absolute timestamp. For more information, see Point-in-time recovery. For details on the yb-admin commands, refer to the Backup and snapshot commands section of the yb-admin documentation.

You can try out PITR by creating a namespace and populating it, creating a snapshot schedule, and restoring from that schedule.

Note

This document contains examples that are deliberately simplified. In many of the scenarios, you could drop the index or table to recover. Consider the examples as part of an effort to undo a larger schema change, such as a database migration, which has performed several operations.

Undo data changes

The process of undoing data changes involves creating and taking a snapshot of a table, and then performing a restore from either an absolute or relative time.

Before attempting a restore, you need to confirm that there is no restore in progress for the subject keyspace or table; if multiple restore commands are issued, the data might enter an inconsistent state. For details, see Restore to a point in time.

Create and snapshot a table

Create and populate a table, get a timestamp to which you'll restore, and then write a row.

  1. Start the YCQL shell and connect to your local instance:

    ./bin/ycqlsh
    
  2. Create a table and populate some sample data:

    create keyspace pitr;
    
    use pitr;
    
    create table employees (
      employee_no integer PRIMARY KEY,
      name text,
      department text,
      salary integer
    ) with transactions = { 'enabled' : true };
    
    insert into employees (employee_no, name, department, salary) values (1221, 'John Smith', 'Marketing', 50000);
    insert into employees (employee_no, name, department, salary) values (1222, 'Bette Davis', 'Sales', 55000);
    insert into employees (employee_no, name, department, salary) values (1223, 'Lucille Ball', 'Operations', 70000);
    insert into employees (employee_no, name, department, salary) values (1224, 'John Zimmerman', 'Sales', 60000);
    
    SELECT * from employees;
    
     employee_no | name           | department | salary
    -------------+----------------+------------+--------
            1223 |   Lucille Ball | Operations |  70000
            1224 | John Zimmerman |      Sales |  60000
            1221 |     John Smith |  Marketing |  50000
            1222 |    Bette Davis |      Sales |  55000
    
    (4 rows)
    
  3. Create a snapshot schedule for the new pitr keyspace from a shell prompt. In the following example, the schedule is one snapshot every minute, and each snapshot is retained for ten minutes:

    ./bin/yb-admin create_snapshot_schedule 1 10 ycql.pitr
    
    {
        "schedule_id": "0e4ceb83-fe3d-43da-83c3-013a8ef592ca"
    }
    
  4. Verify that a snapshot has happened:

    ./bin/yb-admin list_snapshot_schedules
    
    {
        "schedules": [
            {
                "id": "0e4ceb83-fe3d-43da-83c3-013a8ef592ca",
                "options": {
                    "interval": "60.000s",
                    "retention": "600.000s"
                },
                "snapshots": [
                    {
                        "id": "8d588cb7-13f2-4bda-b584-e9be47a144c5",
                        "snapshot_time_utc": "2021-05-07T20:16:08.492330+0000"
                    }
                ]
            }
        ]
    }
    

Restore from an absolute time

  1. Get a timestamp:

    python -c 'import datetime; print(datetime.datetime.now().strftime("%s%f"))'
    
    1620418817729963
    
  2. Add a row for employee 9999 to the table:

    insert into employees (employee_no, name, department, salary) values (9999, 'Wrong Name', 'Marketing', 10000);
    
    select * from employees;
    
     employee_no | name           | department | salary
    -------------+----------------+------------+--------
            1223 |   Lucille Ball | Operations |  70000
            9999 |     Wrong Name |  Marketing |  10000
            1224 | John Zimmerman |      Sales |  60000
            1221 |     John Smith |  Marketing |  50000
            1222 |    Bette Davis |      Sales |  55000
    
    (5 rows)
    
  3. Restore the snapshot schedule to the timestamp you obtained before you added the data, at a terminal prompt:

    ./bin/yb-admin restore_snapshot_schedule 0e4ceb83-fe3d-43da-83c3-013a8ef592ca 1620418817729963
    
    {
        "snapshot_id": "2287921b-1cf9-4bbc-ad38-e309f86f72e9",
        "restoration_id": "1c5ef7c3-a33a-46b5-a64e-3fa0c72709eb"
    }
    
  4. Next, verify the restoration is in RESTORED state (you'll observe more snapshots in the list, as well):

    ./bin/yb-admin list_snapshots
    
    Snapshot UUID                         State
    8d588cb7-13f2-4bda-b584-e9be47a144c5  COMPLETE
    1f4db0e2-0706-45db-b157-e577702a648a  COMPLETE
    b91c734b-5c57-4276-851e-f982bee73322  COMPLETE
    04fc6f05-8775-4b43-afbd-7a11266da110  COMPLETE
    e7bc7b48-351b-4713-b46b-dd3c9c028a79  COMPLETE
    2287921b-1cf9-4bbc-ad38-e309f86f72e9  COMPLETE
    97aa2968-6b56-40ce-b2c5-87d2e54e9786  COMPLETE
    Restoration UUID                      State
    1c5ef7c3-a33a-46b5-a64e-3fa0c72709eb  RESTORED
    
  5. In the YCQL shell, verify the data is restored, and there is no row for employee 9999:

    ycqlsh:pitr> select * from employees;
    
     employee_no | name           | department | salary
    -------------+----------------+------------+--------
            1223 |   Lucille Ball | Operations |  70000
            1224 | John Zimmerman |      Sales |  60000
            1221 |     John Smith |  Marketing |  50000
            1222 |    Bette Davis |      Sales |  55000
    
    (4 rows)
    

Restore from a relative time

In addition to restoring to a particular timestamp, you can also restore from a relative time, such as "ten minutes ago".

When you specify a relative time, you can specify any or all of days, hours, minutes, and seconds. For example:

  • "5m" to restore from five minutes ago
  • "1h" to restore from one hour ago
  • "3d" to restore from three days ago
  • "1h 5m" to restore from one hour and five minutes ago

Relative times can be in any of the following formats (again, note that you can specify any or all of days, hours, minutes, and seconds):

  • ISO 8601: 3d 4h 5m 6s
  • Abbreviated PostgreSQL: 3 d 4 hrs 5 mins 6 secs
  • Traditional PostgreSQL: 3 days 4 hours 5 minutes 6 seconds
  • SQL standard: D H:M:S

Refer to the yb-admin restore-snapshot-schedule command for more details.

Undo metadata changes

In addition to data changes, you can also use PITR to recover from metadata changes, such as creating and altering tables, and creating indexes.

Undo table creation

  1. Using the same keyspace as the previous scenarios, create a new table:

    create table t2(k int primary key);
    
  2. Restore back to a time before this table was created, as described in Restore from a relative time.

  3. Due to a ycqlsh caching issue, to check the effect of this change, you need to drop out of your current ycqlsh session and log back in.

  4. Check that table t2 is gone:

    ./bin/ycqlsh -e 'use pitr; describe tables;'
    
    employees
    

Undo table alteration

Undo column addition

  1. Using the same keyspace as the previous scenarios, alter your table by adding a column:

    alter table pitr.employees add v2 int;
    select * from pitr.employees;
    
     employee_no | name           | department | salary |  v2
    -------------+----------------+------------+--------+------
            1223 |   Lucille Ball | Operations |  70000 | null
            1224 | John Zimmerman |      Sales |  60000 | null
            1221 |     John Smith |  Marketing |  50000 | null
            1222 |    Bette Davis |      Sales |  55000 | null
    
    (4 rows)
    
  2. Restore back to a time before this table was altered, as described in Restore from a relative time.

  3. Due to a ycqlsh caching issue, to check the effect of this change, you need to drop out of your current ycqlsh session and log back in.

  4. Check that the v2 column is gone:

    ycqlsh:pitr> select * from employees;
    
     employee_no | name           | department | salary
    -------------+----------------+------------+--------
            1223 |   Lucille Ball | Operations |  70000
            1224 | John Zimmerman |      Sales |  60000
            1221 |     John Smith |  Marketing |  50000
            1222 |    Bette Davis |      Sales |  55000
    
    (4 rows)
    

Undo column deletion

  1. Using the same keyspace as the previous scenarios, alter your table by dropping a column:

    alter table pitr.employees drop salary;
    select * from pitr.employees;
    
     employee_no | name           | department
    -------------+----------------+-----------
            1223 |   Lucille Ball | Operations
            1224 | John Zimmerman |      Sales
            1221 |     John Smith |  Marketing
            1222 |    Bette Davis |      Sales
    
    (4 rows)
    
  2. Restore back to a time before this table was altered, as described in Restore from a relative time.

  3. Due to a ycqlsh caching issue, to check the effect of this change, you need to drop out of your current ycqlsh session and log back in.

  4. Verify that the salary column is back:

    ycqlsh:pitr> select * from employees;
    
     employee_no | name           | department | salary
    -------------+----------------+------------+--------
            1223 |   Lucille Ball | Operations |  70000
            1224 | John Zimmerman |      Sales |  60000
            1221 |     John Smith |  Marketing |  50000
            1222 |    Bette Davis |      Sales |  55000
    
    (4 rows)
    

Undo index creation

  1. Create an index on the table from the previous examples:

    use pitr;
    create index t1_index on employees (employee_no);
    describe index t1_index;
    
    CREATE INDEX t1_index ON pitr.employees (employee_no)
    WITH transactions = {'enabled': 'true'};
    
  2. Restore back to a time before this index was created.

  3. Due to a ycqlsh caching issue, to check the effect of this change, you need to drop out of your current ycqlsh session and log back in.

  4. Verify that the index is gone:

    ./bin/ycqlsh -e 'describe index pitr.t1_index;'
    
    <stdin>:1:Index u't1_index' not found
    

Undo table and index deletion

  1. Create a second table called dont_deleteme:

    create table dont_deleteme (
      oops integer PRIMARY KEY,
      mistake text
    ) with transactions = { 'enabled' : true };
    
    describe tables;
    
    employees  dont_deleteme
    
  2. Create an index on the table from the previous examples:

    use pitr;
    create index t1_index on employees (employee_no);
    describe index t1_index;
    
    CREATE INDEX t1_index ON pitr.employees (employee_no)
    WITH transactions = {'enabled': 'true'};
    
  3. Wait a minute or two, then delete the new table and index:

    drop table dont_deleteme;
    drop index t1_index;
    
  4. Restore back to a time before you deleted the table and index, as described in Restore from a relative time.

  5. Verify that the index and table are restored:

    ./bin/ycqlsh -e 'describe table pitr.dont_deleteme; describe index pitr.t1_index;'
    
    CREATE TABLE pitr.dont_deleteme (
        oops int PRIMARY KEY,
        mistake text
    ) WITH default_time_to_live = 0
        AND transactions = {'enabled': 'true'};
    
    CREATE INDEX t1_index ON pitr.employees (employee_no)
        WITH transactions = {'enabled': 'true'};