[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Ext3 and external journals...

On Nov 09, 2001  16:32 +0000, Stephen Tweedie wrote:
> On Thu, Nov 08, 2001 at 02:55:25PM -0700, Andreas Dilger wrote:
> > Hmm, a potential problem is that root must be mounted r/w in case
> > the journal device needs to dump the contents for an unknown client
> > filesystem.
> Indeed.  Shared journal for the root device is a sure path to madness:
> we have _got_ to be able to assume a writable root at that point.

No disagreements here.  It may need some tweaks to startup scripts though.

> > Having the journal device in passno=2 means it could
> > be recovered out-of-order with the client filesystem because of
> > device-sharing issues.  An easy solution is to put users of journal
> > devices into passno=3, but it may be desirable to also have some
> > safeguards in case people don't do this correctly (maybe have fsck
> > or fsck.ext3 smart enough to block attempts to fsck the client before
> > a shared journal).
> Possible, but it's an ugly UI wart.  We really ought to be able to
> find the journal, replay the specific components for a given client
> fs, and mark the client and/or journal has being fully recovered for
> that client only.  Even if you don't want to do that at boot time,
> it's important to be able to do it from a rescue boot.

Yes, I was thinking this as well.  Doing e2fsck on a 1:1-external-journal
ext3 filesystem should be the same as doing it on the journal device.

When we get into shared journals, it would be good to be able to do as
you say - fsck only one filesystem, and "invalidate" those transactions
from the journal device.  If you do it on all of the journal clients,
the journal device should be empty.  I guess it is up to the design of
the shared journal to allow partial cleanup of the journal device like
this.  It may be that we need to do this anyways, when fsck'ing the
journal device, just that we implicitly do fsck on all of the clients of
the shared journal at that time.

Cheers, Andreas
Andreas Dilger

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]