Hi Jakob,
Sorry for not replying earlier. Missed reading your email completely..:(
PVFS2 should work with any conventional backup software (tivoli, tar
gzip) unlike PVFS1
The FAQ sheds some light on this..
http://www.pvfs.org/pvfs2-faq.html#SECTION000101000000000000000

If your meta data is stored on shared storage, you could configure
failover with Linux-HA and have the failed manager taken over by
another manager.
If it is stored on a local disk and if the disk is damaged (and you
don't have a way of rebuilding it such as with RAID 5) all the MD on
that drive is gone unfortunately..
One workaround that is to configure DRBD to emulate a shared storage
over TCP/IP.
Then format a file system and pvfs2 volume on top of that...
http://www.parl.clemson.edu/pipermail/pvfs2-developers/2005-June/001310.html

If one I/O node is down, only 1/10th of your file data will be
unavailable until the node
comes back online. If its disk is toast, then 1/10th of all file's
data will be lost I am afraid since we only do striping.
If you only care about availability, then I guess we need to add some
extra code/functionality to mark that certain servers are off-line to
remember not to make use of them for future I/O or meta data
operations.. This should be feasible to do, I think..

Hope this answers your questions,
thanks,
Murali

On 1/15/07, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote:
hi
I am pvfs 1 user but this question in relevant also to pvfs 2.
I want to know if there are backup options.
for example what would happen if my mgr server is damaged.
can i rebuild the files?

another scenario: if my topology contain 10 i/o nodes and each file is strip
between all
the 10 nodes.
if one node is damaged does all my data will be lost?

thanks.
Jakob.
_______________________________________________
Pvfs2-users mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users



_______________________________________________
Pvfs2-users mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users

Reply via email to