On Mar 2, 2010, at 1:59 PM, Matt wrote:> So I''m in the process of building a ZFS based SAN. After toying
with it at home I''ve ordered up all the parts to begin my build.
That''s a completely different story though.
>
> I''m wondering what the possibilities of two-way replication are
for a ZFS storage pool.
>
> The scenario - the ZFS SAN will be used as an NFS mount (hopefully) for
Virtul Machine storage over Infiniband. What I''d like to be able to do
is replicate this to a second datacenter with the ability to fail over to the
second datacenter.
>
> I''m well aware of being able to ship snapshots over SSH, and I
think I''ve got a big enough datacenter link that I could do this to the
second datacenter without much problem. The concern comes in as soon as I
initiate a failover.
>
> Once I fail to the second system, all changes are going to the second
system, while the primary system is offline/unavailable. When the primary
system comes back online, the secondary system will not really have (to my
knowledge) any idea what state the primary system is in. Is there a way to then
push any changes made locally on the second system back to the primary system
without doing a full replication?
Good question. If you use send/receive, then the increments are snapshots.
It doesn''t matter what system is the source of the snapshots, but the
destination
cannot have data added after the last, common snapshot. To reverse the flow,
just send from the secondary to the primary. I''m currently working on
a slide to
show this in pictures, but until then does this make sense?
> Most of the cutover can be handled manually, and the second system could
live in an "offline" state where the data isn''t writeable
until I tell it to be. 20-30 minutes of downtime while I power up VM''s
is acceptable. My biggest concern is that when the primary system comes back
online that I don''t have to push 20TB+ back to the primary system,
especially if it was only offline for 2 hours.
>
> I''ve looked in to the HA Cluster, but for some reason my login
information won''t allow me to look at it currently.
Seems to be MIA for me, too :-(
You should check out NexentaStor''s cluster as another solution.
Nexenta''s
snapshot schedules and remote replication are easy to setup and manage.
-- richard
ZFS storage and performance consulting at http://www.RichardElling.com
ZFS training on deduplication, NexentaStor, and NAS performance
http://nexenta-atlanta.eventbrite.com (March 16-18, 2010)