[Pvfs2-users] Fail if one i/o node is gone
Robert Latham
robl at mcs.anl.gov
Tue Dec 2 11:10:08 EST 2008
On Tue, Dec 02, 2008 at 04:53:15PM +0100, Michael A. Toth wrote:
> On Tue, Dec 2, 2008 at 4:44 PM, Emmanuel Florac <eflorac at intellique.com> wrote:
> >
> > That's easy in principle : you'll have to create the code that does
> > that. I personally thought about it and came to the conclusion that it
> > may be feasible only by modifying the pvfs2-client; however it's far
> > from easy. And I still wonder how to achieve metadata redundancy.
>
> Hi Emmanuel,
>
> Ahh, I see, thanks. I have to find an another distfs solution.
You might not have to give up on PVFS just yet, but it will take more
hardware.
You can set up PVFS servers in an active-active high
availability config, and then when one server goes down, another
machine will take over for it. This setup does require shared
storage.
Take a look at the pvfs2-ha and pvfs2-ha-heartbeat-v2 documents (in
the doc/ directory) for more information.
==rob
--
Rob Latham
Mathematics and Computer Science Division A215 0178 EA2D B059 8CDF
Argonne National Lab, IL USA B29D F333 664A 4280 315B
More information about the Pvfs2-users
mailing list