Currently to have redundancy you need to build your system with a drbd or
similar replicated block device in conjunction with heartbeat or corosync.
 We are developing toward v3 which will have metadata and file data
redundancy built into the FS layer and not have to rely on lower level
redundancy.

For larger installations people have built a few redundant MD servers, then
when an IO or file Data server is unavailable new writes would work, only
reads from the failed system would fail.   As of 2.8.6 there is also the
ability to have replication for immutable files (mainly for content
delivery).  But the full redundancy will be in v3.

Ps. The multi-server configuration with distributed MD and FD is for
performance, as you add nodes you get better performance because the files
can be striped and leveraged in conjunction with the PVFS protocol that can
work with multiple servers concurrently, essentially distributing the load

-boyd


On Wed, Apr 2, 2014 at 11:53 AM, Victor Belizário <
[email protected]> wrote:

> Hey guys,
>
> I successfully completes the installation of PFVS2 on two CentOS servers,
> both playing the roles of master, storage and client. Each pointed to
> himself as "mount-t pvfs2 tcp ://localhost:3334/pvfs2-fs /mnt/orangefs" and
> everything went right, the communication is perfect and what I create on a
> server is automatically created for another server. My question is: I'm
> testing now the failover, purposely dropped a server (ifdown eth0) but when
> i do that the other server stopped responding and was unable to access over
> the mount point /mnt/orangefs. Should continue working, correct? Even with
> one of the other server offline the other server should continue operating
> normally or not? Thank you!
>
> _______________________________________________
> 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