Was 40 days, this reduces it to 21 days. That should roughly halve
the amount of disk space needed to store all the WALs, at the cost
that it's more difficult to recover a replica which has been out of
sync for between 22 and 40 days.
Closes #121
#!/bin/sh
-# Cleanup archive directory
-find -L /store/postgresql/archive -mtime +40 -print0 | xargs -0r rm
+# Cleanup archive directory - keep 3 weeks of WALs
+find -L /store/postgresql/archive -mtime +21 -print0 | xargs -0r rm