A vanilla installation of DSpace is configured to log all activity to [dspace]/log/dspace.log. Once the log file reaches 1mb, it renames itself to dspace.log.1, and DSpace writes to a new dspace.log. Once this fills up, it renames dspace.log.1 to dspace.log.2, renames dspace.log to dspace.log.1, and starts a new dspace.log. This continues until dspace.log.500 is reached, upon which any subsequent log rolling will cause dspace.log.500 to be deleted, and the data is lost.
A downside of this is that backup systems will back up all log files every night, as the filename for a given log file will change every time the logs roll over. This can make it hard to recover old log files from a backup system.
The dspace.log file performs several functions:
- It provides log information useful for debugging technical problems
- It provides some level of access logging, showing who has access the system
- It is used to generate the DSpace statistics pages.
Many sites find that they fill up their 500mb of logs very quickly, often in a matter of days or weeks in a busy site. After that, log files are lost, along with the associated visitor statistics.
One solution is to increase the number of log files held, or increase their size in log4j.properties.
An alternative and more robust solution would be to change the default log4j.properties file to use the DailyRollingFileAppender which will instead create date-stamped files, one per day. No log files will then be lost, and it can be left to the system administrator to manage any deleting of log files.
Please add a comment to this issue, or use the 'Vote' button on the left, if you would like this change to be made in the default DSpace out-the-box configuration.