On Thu, Jan 03, 2008 at 11:21:57AM -0500, Ambrose Li wrote:
> On Thu, Jan 03, 2008 at 10:31:27AM +0100, Petter Reinholdtsen wrote:
> >
> > Also, if you want serialized fsck, I believe it is possible to
> > implement today by specifying differnet sequence numbers in the sixth
> > field of /etc/fstab fs_passno). It specifies the fsck order.
>
> Awesome. How could I have forgetten about that? Thanks!
Serializing fsck will of cours possibly slow down the system
significantly, since you're not taking advantage of the parallelism of
having multiple disk spindles.
Going back to the original bug report:
>However, from fsck's manpage, fsck only displays one progress bar at a
>time even when two checkers are run.
>
>This leads to the following possible scenario where fsck may appear to
>be frozen while it is actively checking a filesystem:
>
>- The system contains at least one very large partition.
>- At startup time, fsck starts to check all partitions (say because the
> box had a very long uptime)
>- fsck chose to run two checkers, one on a small partition and one on
> the big partition
>- fsck chose to display the progress bar on the small partition
>- fsck finish checking the small partition and continues to check all
> the other smaller partitions
>- fsck continues to check the large partition WITHOUT ANY PROGRESS BARS
>
>If the large partition is so large that it takes, say, an hour to finish,
>the machine will appear to be hung.
Have you actually *seen* this happening, or are you positing this as a
theoretical situation just from reading the fsck man page. In fact,
the way fsck works is that when a child e2fsck finishes, fsck will
look for another ext2/ext3 filesystem check taking place, and send
that child e2fsck an USR1 signal, which requests it to start
displaying a progress bar.
So in fact, the right thing should be happening today, to avoid the
problem which you described above.
Regards,
- Ted
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
