CFS file system

[Shivkumar]

Hi,

I read that hpux 11.23 supports CFS file system. It is capable of concurrent file system access.

Can someone explain how this CFS is different than other existing file system on HPUX ?

Thanks,
Shiv

[A. Clay Stephenson · Accepted solution]

Yes. If you try to conventionally mount the same filesystem read-write on more than one host, absolute chaos will result because ServerA will know nothing about ServerB's buffer cache. The Clustered File System allows multiple nodes to mount the same filesystem read-write. To any host, the filesystem appears to be a conventional vxfs filesystem. It does require a dedicated network for the filesystem data.

[Arunvijai_4]

Hi Shiv,

This link should give you more information about CFS http://www.beedub.com/clusterfs.html

-Arun

[Raj D.]

Shiv,

You can also check the below one,:

http://www.clusterfs.com/

Cheers,
Raj.

[Chauhan Amit]

The Veritas CFS is an extension of the VERITAS File System (VxFS). The Veritas CFS allows the same file system to be simultaneously mounted on multiple nodes. Veritas CFS is designed with master/slave architecture. Any node can initiate a metadata operation (create, delete, or resize data) and the master node carries out the actual operation. All other (non metadata) I/O goes directly to the disk.

A distributed locking mechanism, called the global lock manager (GLM) is used for metadata and cache coherency across the multiple nodes. GLM provides a way to ensure that all the nodes will have a consistent view of the file system. When any node wishes to read data, it requests a shared lock. If another node wishes to write to the same area of the file system, it must request an exclusive lock. The GLM revokes all shared locks before granting the exclusive lock and informs reading nodes that their data is no longer valid.

CFS is used in DBE/AC to manage a file system in a large database environment. When used in DBE/AC for Oracle-RAC, Oracle accesses data files stored on CFS file systems with the ODM interface. This essentially bypasses the file system buffer and file system locking. This means that only Oracle handles the tasks of buffering data and coordinating writes to files and not the GLM, which is minimally used with the ODM interface.

Hope this helps.

-Amit

[Shivkumar]

Hi,

Other than Oracle 9i or 10g RAC; how this feature of the CFS will be useful for the normal applications like Apache Web Server or Application Servers like WebLogic and WebSphere etc ?

Thanks,
Shiv

[A. Clay Stephenson]

Potentially any application could benefit from this -- but the application would still have to deal with locking contention. Without CFS, the only way multiple hosts could access the same physical disks safely was by using raw i/o; ie - no filesystem. Applications could also use NFS to read and write to a common filesystem but with significant performance impacts. With CFS, multiple hosts can read and write to the same locally attached filesystem using fully cooked i/o.

The application must be designed to handle this. Consider 2 users vi'ing the same file on the same host. Because no locking is done by the application (vi in the case) whoever writes to the file last, wins. Vi would have exactly the same problem under CFS when the users were on separate systems.

CFS should be thought of as a new tool to put in the designer's toolbox. What was only possible before with raw/io is now possible with cooked files.

[Shivkumar]

Hi Clay,

If 2 users changing same vi file under CFS on the same host; then, will both changes will go through successfully ?

Regards,
Shiv

[A. Clay Stephenson]

Yes. Both writes will succeed but the last write will overwrite the entire file. Vi is not designed to handle locking so it will work equally well (or equally badly) under a local vxfs filesystem or a CFS filesystem. CFS is designed to be invisible to the application so if an application correctly handles lock contention as a local file, it will also correctly handle lock contention as a CFS file.

File locking in UNIX is advisory only (although it is possible to promote advisory locks to mandatory locks). This means that the application itself must be designed to both lock files (or regions of a file) and check for locks on files.

[inventsekar_1]

i am not good with CFS and all.
but i can find some good links from that u can read something good.


Cluster file system from wikipedia:
http://en.wikipedia.org/wiki/Cluster_%28file_system%29

cfs faq:
http://www.clusterfs.com/faq.html

Selecting a Scalable Cluster File System:
http://www.lustre.org/docs/selecting-a-cfs.pdf

CFS roadmap:
http://www.clusterfs.com/Lustre-roadmap-2006-05-22.pdf

Using Serviceguard Extension for RAC from docs.hp.com
http://docs.hp.com/en/T1859-90038/ch01s03.html