C.3.35. Starting replication after performing a restore because of an invalid restart sequence number

Last Updated: 2013-11-01

Condition or Error

Starting replication fails because of an invalid restart sequence number. Checking the sequence number, trep_commit_seqno shows an empty or invalid table contents:

mysql> select * from tungsten.trep_commit_seqno;
+-------+--------+-----------+-----------+--------------+---------+-----------------+---------------------+
| seqno | fragno | last_frag | source_id | epoch_number | eventid | applied_latency | update_timestamp |
+-------+--------+-----------+-----------+--------------+---------+-----------------+---------------------+
| -1 | NUL L | NULL | NULL | NULL | NULL | -1 | 2013-10-27 23:44:05 |
+-------+--------+-----------+-----------+--------------+---------+-----------------+---------------------+
1 row in set (0.00 sec)

Causes

  • The restore may have failed to correctly restore the tungsten tables.

Rectifications

  • Retry the restore process, making sure the replicator is stopped and there are no updates to the table taking place:

    1. Ensure no replicator processes are running:

      shell> replicator stop

      Ensure the /opt/continuent/thl directory is empty:

      shell> tpm reset-thl 
    2. Restore the backup using whatever backup/restore tool you used.

    3. Check that trep_commit_seqno has a valid restart position in it

    4. shell> replicator start