DragonFly BSD
DragonFly users List (threaded) for 2006-08
[Date Prev][Date Next]  [Thread Prev][Thread Next]  [Date Index][Thread Index]

Re: Question about journaling


From: Matthew Dillon <dillon@xxxxxxxxxxxxxxxxxxxx>
Date: Fri, 18 Aug 2006 09:10:47 -0700 (PDT)

:Hi,
:
:I've read the mountctl(8) and jscan(8) manpages, but
:there's one thing which I'm still not 100% certain.
:
:Does the journaling work on file level or on block
:level?  Specifically:  When I use jscan to mirror a
:file system on a remote machine, will the files have
:identical inode numbers?
:
:That's important for setting up a fail-over NFS cluster
:(with the master being read+write, and the slave being
:read-only, replicated with jscan).  If the inode numbers
:are different, a transparent fail-over would not be
:possible ...  In that case, is there are different
:solution with DragonFly BSD?
:
:Best regards
:   Oliver
:
:-- 
:Oliver Fromme,  secnetix GmbH & Co. KG, Marktplatz 29, 85567 Grafing
:Dienstleistungen mit Schwerpunkt FreeBSD: http://www.secnetix.de/bsd

    The code runs at the file level.  The inode number is recorded 
    in the stream but the mirroring code has no ability to specify
    the inode when creating new nodes on the target so the inode
    number is not replicated.

    Apart from that, there are still issues with any disruption that
    might occur to the journaling stream.  For example, if a connection
    between two machines is lost.  I think the restart code should work,
    but it isn't as robust as I would like because the source filesystem
    could end up blocking indefinitely if the journaling stream cannot be
    reestablished.

    I consider the journaling code to still be quite experimental.  It is
    more a placemark on how we might have to encode data updates in a
    clustered system then it is an actual journaling system.

						-Matt




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