High Availability Shared File System Recommendation

• Previous message: Symantec Veritas: "NFS stale file handles"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Date: Thu, 24 Feb 2005 19:41:29 -0800
To: sunmanagers@sunmanagers.org

Hello,

I am currently trying to figure out what is the best file system to
use in a highly available system. I would appreciate any advise
anyone can provide and will summarize upon completion.

A little background first, I am configuring a system with 4 Sun Fire
240s and 1 fully populated Sun StorEDGE 3510. This is going to be a
web based system with 2 -240s serving as a load balanced/clustered
web/application server and the other 2 240s will be Oracle 10g
clustered with RAC. The 3510 is directly attached to the 240s and
will house files that need to be read/write from the application
server as well as the Oracle CRS, ASM and Datastore. The Operating
System is Solaris 8. Performance of the system is not extremely
critical, however uptime and reliability is. The mean time to recover
the system to full functionality has been set at 5 minutes. This
configuration has not been set in stone and is somewhat flexible.

The application server and Oracle RAC have been clustered
successfully. However, if the machines containing the mount points to
the 3510 fail, obviously the application and database becomes
unavailable.

I would like to provide a mount point where read/write access is
provided through a shared files system to the the 3510 from each 240
for its respective purpose.

Possible solutions being discussed are the following:
1. NFS but what happens if the server running the NFS server fails?
How difficult is it to write a script to monitor a heartbeat of the
NFS server and perform failover if failure is detected? Are there any
scripts already available to do this?
2. QFS I understand this is really for performance and doesn't really
address reliability. Is this true? When enabling the shared file
option doesn't it have the same type of issues as NFS for reliability,
what if the metadata server fails?
3. Sun Cluster, I understand this can do the fail over and sharing of
file system through NFS and/or QFS. But, is this complexity really
needed?
4. 3rd party clustered files system, like Veritas. What issues exist
here? What components are really needed?
5. NAS device, but isn't this really shifting the burden of failure to
the NAS. What is the likelihood of a NAS device failing?

I am leaning towards trying to automate an NFS solution but as you can
see I am somewhat unfamiliar with file systems and would appreciate
any advise, links to whitepapers, as well as success/failure stories.

Thanks in Advance,
-Andrew
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers

• Previous message: Symantec Veritas: "NFS stale file handles"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]