Force a checkpoint

When necessary, you can force a checkpoint with an onmode or SQL administration API command.

Force a checkpoint in any of the following situations:
  • To free a logical-log file that contains the most recent checkpoint record and that is backed up but not yet released (onstat -l status of U-B-L or U-B)
  • Before you issue onmode -sy to place the database server in quiescent mode
  • After building a large index, if the database server terminates before the next checkpoint. The index build restarts the next time that you restart the database server.
  • If a checkpoint has not occurred for a long time and you are about to attempt a system operation that might interrupt the database server
  • If foreground writes are taking more resources than you want (force a checkpoint to bring this down to zero temporarily)
  • Before you run dbexport or unload a table, to ensure physical consistency of all data before you export or unload it
  • After you perform a large load of tables using PUT or INSERT statements (Because table loads use the buffer cache, forcing a checkpoint cleans the buffer cache.)

To force a checkpoint, run onmode -c.

Alternatively, if you use ON-Monitor (UNIX™), choose the Force-Ckpt option from the main menu. The time in the Last Checkpoint Done field does not change until a checkpoint occurs. The Last Checkpoint Check field shows the time of the last checkpoint check. If no modifications have been made since the time of the last checkpoint, the database server does not perform a checkpoint.

For information about using SQL administration API commands instead of some onmode commands, see Remote administration with the SQL administration API and the HCL OneDB™ Guide to SQL: Syntax.