Why does filesystem gradually increase by mlocate.db database ?

Environment

• Red Hat Enterprise Linux 4
• Red Hat Enterprise Linux 5
• Red Hat Enterprise Linux 6

Issue

• Have many mlocate.db.xxxxxx files in /var/lib/mlocate. The filesystem is currently full.
• The date of mlocate.db is old, with the "extra" files all newer.
• Need to know if it is safe to remove these files and run updatedb manually to free up disk space in /var.

Resolution

• It's safe to delete mlocate.db.xxxxxx files in /var/lib/mlocate , they'll just get created from scratch next time mlocate runs.
• If nobody is using locate command, then it can be disabled in /etc/cron.daily (mlocate.cron) so that it can avoid from the issue.
• If the /var/lib/mlocate.db.<xxxxxx> files are all newer than the actual /var/lib/mlocate.db file, then the daily cron job is likely either being interrupted and killed or is hanging and never getting a chance to write back the temporary working file to the /var/lib/mlocate.db location.
• Example strace of an updatedb:
  • When updatedb starts, it open a temporary working file ...

8242   open("/var/lib/mlocate/mlocate.db.hwCz3x", O_RDWR|O_CREAT|O_EXCL, 0600) = 4

• Once it's finished, it moves it's temporary file back into the original location

8242   chown("/var/lib/mlocate/mlocate.db.hwCz3x", 4294967295, 21) = 0
8242   chmod("/var/lib/mlocate/mlocate.db.hwCz3x", 0640) = 0
8242   rename("/var/lib/mlocate/mlocate.db.hwCz3x", "/var/lib/mlocate/mlocate.db") = 0