Maintaining the FireWorks Database

There are two types of maintainence operations that can be performed through the LaunchPad:

  • maintain - automatically detects and marks failed jobs and deleted queue reservations
  • tuneup - tries to speed up database performance by updating indices and compacting the database. Should be performed during database downtime.

Maintain to mark failed jobs and queue reservations

We recommend that all users periodically run the maintain command from the LaunchPad. This will detect failed jobs and queue reservations - for details, refer to the failures tutorial and queue reservation tutorial. In summary, the command will mark the state of failed jobs as FIZZLED and will move any long-waiting RESERVED FireWorks back to READY. The maintenance command can be run using:

lpad maintain

If you would like to run maintenance in an infinite loop, you can use:

lpad maintain --infinite --maintain_interval 6000

This formulation will run a maintenance job every 6000 seconds (100 minutes).

Tuneup to improve performance

You only need to run a database tuneup if you are not satisfied with the performance of FireWorks. The tuneup will update all the indices and compact the database. You should only run the tuneup command during periods of downtime. You can run tuneup using:

lpad tuneup

Table Of Contents

Previous topic

Dealing with Failures and Crashes

Next topic

Querying FireWorks and Workflows

This Page