That works. Great, thanks. How do you test it using the VFS code path anyway?
John On Wed, 23 Nov 2005, Murali Vilayannur wrote: > Hi John > > Does the attached patch fix it? > It was a mistake in the code (not in the way you are using it). I only > tested the VFS code path and not the system interface path. Thanks for > the reports! > > > pascal006:~>pvfs2-viewdist -f /mnt/pvfs2/hosts > > Not a meta file! > > Could not open /mnt/pvfs2/hosts > > > > The file does exist though: > > pascal006:~>pvfs2-ls /mnt/pvfs2/hosts > > hosts > > > > Am I doing something funny? > > > > Btw, the usage for pvfs2-viewdist is incorrect. It says: > > Usage: pvfs2-viewdist ARGS src_file dest_file > > where ARGS are blah. > > But the code only looks at ARGS and not for src_file and dest_file. Must > > have been copied from pvfs2-cp. :) > > Oops. How embarassing :) I fixed that in the attached patch.. > > > Thanks by the way for all of the speedy replies I've been getting. It > > really helps a lot to have such a well-monitored and well-supported > > mailing list. > > Thank you for all reports! > Murali > > > > > John > > > > On Mon, 21 Nov 2005, Murali Vilayannur wrote: > > > > > Hi, > > > Could you give the attached patch a go and see if it works out? > > > All it does is print the servers that hold chunks/strips of a file (both > > > for the VFS and system interface users) using xattrs. > > > Patch is against CVS. > > > P.S: RobR, I noticed that we arenot setting the fs_id field of the > > > statfs structure (statfs/fstatfs system call). Is/are there any > > > reason/reasons for not doing so? > > > I need that change to have VFS clients know the fs_id. > > > Thanks > > > Murali > > > > > > On Mon, 21 Nov 2005, Hamza KAYA wrote: > > > > > > > Hi, > > > > Thank you for the code. It works great:) I was able to find the > > > > distribution parameters of a file which was created by varstrip-dist. > > > > So by setting the jitter variable to a constant value in > > > > PINT_cached_config_get_next_io, I'll be able to find the physical > > > > mapping of a file to the datanodes, at least I hope so:) am I wrong? > > > > > > > > Thanks very much. > > > > > > > > P.S.: > > > > I posted this mail directly to Murali. So I re-post it to the list. > > > > Sorry for the inconvenience. > > > > > > > > -- > > > > Hamza > > > > > > > > On 11/18/05, Murali Vilayannur <[EMAIL PROTECTED]> wrote: > > > > > Hi, > > > > > Does the attached program tell you the distribution at least? Compile > > > > > it > > > > > like you compile pvfs2-cp or some other program. It uses the VFS > > > > > fgetxattr > > > > > system call or the PVFS2 pvfs_sys_geteattr. > > > > > If it works, then I will work on it some more and print out more of > > > > > the information that you wanted... > > > > > Thanks, > > > > > Murali > > > > > > > > > > On Fri, 18 Nov 2005, Hamza KAYA wrote: > > > > > > > > > > > Hi, > > > > > > > > > > > > Using PVFS2 system calls I was able to apply the varstrip > > > > > > distribution. Using pvfs2-fs-dump I can see the results. But I can't > > > > > > find out the reverse operation. I mean how can I learn the > > > > > > distribution of a file or the physical location of a given file > > > > > > using > > > > > > system calls? > > > > > > > > > > > > -- > > > > > > hamza > > > > > > > > > > > > _______________________________________________ > > > > > > PVFS2-users mailing list > > > > > > [email protected] > > > > > > http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > _______________________________________________ > > > > PVFS2-users mailing list > > > > [email protected] > > > > http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users > > > > > > > > > > > > _______________________________________________ PVFS2-users mailing list [email protected] http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users
