Re: [SSI] CFS status

[Aneesh K. K.V <ane...@di...>]

Hi,


On Fri, 2002-05-24 at 07:23, David B. Zafman wrote:
> 
>
[snip.....]
> Future work:
> ------------
> Handle nodes joining a cluster with existing CFS mounts that are not the 
> root filesystem
> 
> Verify node down handling
> 
> Support file locking posix and flock across the cluster (no mandatory)
> 
> Add HA (high availability) support.  Failover CFS server for disks which 
> can be accessed from multiple nodes.
> 
I think once we have this we can have a I/O balancing ( mentioned in my
previous mail regarding remount ) because i can make use of the fail
over  code path to associate another node as the CFS server for some
partition. ie,  if i have a configuration like 

 		node1			node2
		 |			 |
		 |			 |
		  ------------------------
			   |
			   |
			---------
			|storage |
			--------- 
Now node1 is the master it comes up with the images store in the local
hardisk and make the storage as the root partition. ( The storage here
is not GFS it could be any other file system.ext3, resierFS... ) and it
is mounted as a CFS. that means node1 servers all the request for files
in the "sorage". Now when node2 comes up either through ethernet booting
or booting from a local hardisk  it mounts the CFS root . After that i
can initiate a "failover"  sort of operation for /home partition inside
the "storage"  so that node2 becomes the CFS server for /home.

 Is it possible ? 

 -aneesh