[SAGE] Round #2: no more NFS please

[g.ennis@ns1.adnog.org]

Hi all

The senario:

"Production" runs from two Data Centers.
99% uptime required.

Current config:  NFS running as a "package" under MC/ServiceGuard. If Data 
Center A goes down, the NFS package is "relocated" (via MC/ServiceGuard) to 
Data Center B.  With the MC/ServiceGuard package, goes it's IP address. Hence a 
bridged subnet between Data Centers A & B. (Ip address has to make sense in 
both Data Centers)

So now - if management says: "The bridged subnet has to go". You lose the 
bridged subnet - you lose the MC/SericeGuard fail over ability for NFS. So, no 
more NFS dependance allowed in production, because no fail over.

Question:
Is there any other application/filesystem type, that could provide for the 
sharing of files/file systems ( now across subnets). Hosts in both data centers 
would have to have access to shared files and in the event of the complete loss 
of one data center, the service must keep running and serving files.

Let me give a classic example. Currently - one NFS mount /transfer is used for 
the push & pull of data between apps. AppOne runs on hostA1 (data center A) and 
on hostB1 (data center B) both instances of AppOne processes data out 
of /transfer. If either of the two has a problem, processing continues 
unaffected at a lesser throughput. Whether the problem was small and limited to 
hostA1 or huge and took out the entire data center A. 

Any thoughts or suggestions?? Given the +99% uptime requirement - whatever 
replacement is suggested, it would have to be very stable and robust. 
Nothing "experimental" is going to be acceptable. The preference would be for 
something one could purchase 24x7 support for.

All suggestions, thoughts and comments gratefully accepted. If you want to 
reply off-list I will summarize.

Gonda