[Linux-cluster] Fwd: High Available Transparent File System

Digimer linux at alteeve.com
Sun Apr 10 14:57:45 UTC 2011


On 04/10/2011 10:29 AM, Meisam Mohammadkhani wrote:
> Hi All,
> 
> I'm new to GFS. I'm searching around a solution for our enterprise
> application that is responsible to save(and manipulate) historical data
> of industrial devices. Now, we have two stations that works like hot
> redundant of each other. Our challenge is in case of failure. For now,
> our application is responsible to handling fault by synchronizing the
> files that changed during the fault, by itself. Our application is
> running on two totally independent machines (one as redundant) and so
> each one has its own disk.
> We are searching around a solution like a "high available transparent
> file system" that makes the fault transparent to the application, so in
> case of fault, redundant machine still can access the files even the
> master machine is down (replica issue or such a thing).
> Is there fail-over feature in GFS that satisfy our requirement?
> Actually, my question is that can GFS help us in our case?
> 
> Regards

Without knowing your performance requirements or available hardware, let
me suggest:

DRBD between the two nodes
GFS2 on the DRBD resource.

This way, you can use DRBD in Primary/Primary mount and mount the GFS2
share on both nodes at the same time. GFS2 required DLM, distributed
lock manager, so you will need a minimal cluster setup. To answer your
question directly; GFS2 does not need to fail over as it's available on
all quorate cluster nodes at all times.

If you just want to ensure that the data is synchronized between both
nodes at all times, and you don't need to actually read/write from the
backup node, then you could get away with just DRBD in Primary/Secondary
mode with a normal FS like ext3. Of course, this would require manual
recovery in the even of a failure, but the setup overhead would be a lot
less.

If either of these sound reasonable, let me know and I can help give you
more specific suggestions. Let me know what you have in way of hardware
(generally; NICs, Switches, etc).

-- 
Digimer
E-Mail: digimer at alteeve.com
AN!Whitepapers: http://alteeve.com
Node Assassin:  http://nodeassassin.org




More information about the Linux-cluster mailing list