https://bugzilla.novell.com/show_bug.cgi?id=758391 https://bugzilla.novell.com/show_bug.cgi?id=758391#c12 --- Comment #12 from Leonardo Chiquitto <lchiquitto@suse.com> 2012-11-12 11:01:15 UTC --- There's one fundamental difference between the AutoFS kernel modules from 11.4 and 12.1. This might explain why you're seeing the behavior described here. Some back ground information: Every AutoFS mount point has a timeout defined. This timeout specifies how long AutoFS will wait before trying to unmount the volume. The default timeout in our configuration (/etc/sysconfig/autofs) is 600 seconds, or 10 minutes. Every $TIMEOUT/4 seconds, AutoFS checks whether the volume is being used or not and updates the expiration timer accordingly, lets call this operation the "expiration check". This is the important difference: On kernels older than 2.6.39, the expiration check is basically asking the kernel whether the mount point can be unmounted *now*. If the kernel says 'no', it means that at least one process holds an open file in or has the current working directory set to that mount point. In this case, when the volume is effectively busy, AutoFS will reset the expiration timer. On kernels 2.6.39 and newer, the expiration check is much broader: it knows exactly when the volume was last accessed. In other words, the volume doesn't need to be busy at the time of the check for the timer to be reset: every file accessed causes a reset in the timeout. A practical example: Suppose /nfs is an automounted volume with a timeout of 60 seconds and that /nfs/file is a regular text file. If you run the following command: $ while true; do cat /nfs/file; sleep 5; done .. on openSUSE 11.4 (kernel 2.6.37), you'll notice that ~60 seconds after /nfs was mounted, it will be unmounted, regardless of the multiple accesses to /nfs/file. Few seconds after the unmount, 'cat' will trigger the mount again. .. on openSUSE 12.1 (kernel 3.1), you'll notice that as long as the command is running, the volume will NOT be unmounted. Bottom line is: on 12.1 expiring a mount point is, in general, less likely to happen. Expiration only happens when no access to files or directories inside the mount point happened during $TIMEOUT seconds. Please check if you don't have processes accessing the volume in a frequency shorter than the timeout. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.