Some trace files will have errors or trace information from unsolved problems; others may have performance statistics from some recent test runs. If there is a need to keep these files, they should be filed in separate named directories, where the housekeeping scripts will not find and delete them. Ideally, such files should be transferred off the live data warehouse machine onto the development machine.

Some trace information is worth keeping in its own right. For example, some RDBMSs such as Oracle generate a log file that contains all major operations run against the database. These operations include startup and shutdown, addition or deletion of data files, changes to journal files and so on. These are worth filing as a history log. The logs themselves can become too big to manage easily or browse, so they should be backed up and then cleared down.