[email protected] wrote: > Of course they are not. I'm simply trying to understand the capabilities > and limitations of the Oracle cluster filesystem. With an NFS mounted > filesystem, you can have two servers mount the same volume and perform > simultaneous reads/writes while the NFS mechanism takes care of locking, > etc. This scenario works fine for a mailserver.
Really? What kind of locking are you referring to here? As in, explicit unix locking, flock() or lockf() or something more. > My question is can two non-clustered servers, or two non-clustered > applications, mount the same OFCS2 volume and perform simultaneous > reads/writes without fear of data corruption on the volume? If that application spawns multiple processes, ocfs2 will allow them to run on different nodes and access the same files. Now as far as data sanity goes, it depends on how the application serializes the access. Standard unix mechanisms are flock() and/or lockf(). OCFS2 has support for clustered flock() in OCFS2 1.4 and in mainline kernels, 2.6.26 and higher. While support for lockf() has been added in 2.6.27, the user space cluster stack support required is still a work in progress. So the answer is, it depends on the application. Specifically, it depends on how the application serializes access to the data. _______________________________________________ Ocfs2-users mailing list [email protected] http://oss.oracle.com/mailman/listinfo/ocfs2-users
