Re: [RFC] ext3 multi-mount protection

Hi,

On Fri, 2006-07-14 at 08:26 -0600, Andreas Dilger wrote:

> > secondly, we need it for e2fsck and tune2fs just as much as for mount.
> 
> By all means yes, e2fsck and tune2fs (to a lesser degree, since it is
> generally safe to run on a mounted fs)

tune2fs is generally safe on a *locally* mounted fs.  Remotely mounted
is a different matter altogether; tune2fs will end up doing a non-atomic
read-modify-write which can interfere with superblock writes on the
other node.  It's not likely to be a long-lasting problem though, as the
next sb write will restore things; but even then it does mean that the
tune2fs update will be lost unexpectedly.

e2fsck is worse because it not only needs to be able to test that the fs
isn't mounted, it also needs to mark the fs as in-use while it's
operating, and that requires it to be running the sequence-kicker in the
background.

Something like the exclusive-open flag in cLVM sounds like the right
sort of layer to have this at.  But yes, there's no denying that a
cluster infrastructure is something heavyweight enough that it would be
nice to be able to avoid it.

--Stephen




-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642