Re[4]: [PVFS-users] Re: ROMIO on PVFS

[Rob Ross]

Lei,

Please let us know if the newer pvfs-1.6.2-01292004.patch helps with this.  
Not sure what is going on.

On Tue, 20 Jan 2004, Lei@ICS wrote:

> RR> Is this all on the same node?
> 
> There were two ways to create that large file:
> 1) It was written from a Fortran sequential
> program using the Fortran open(). I don't know
> if in this case PVFS would write the entire file
> to one node.

No, it would be striped with default parameters, which should put it on 
all nodes.

> 2) It was written from a MPI Fortran program
> using MPI_FILE_OPEN(), MPI_SET_VIEW(), MPI_FILE_WRITE().
> This time the file should be written to different
> nodes.
> 
> For both files created in the above two ways,
> ls *.slc does not work.
> 
> RR> Does the file show up in just "ls -l"?
> 
> Yes.
> 
> RR> Is it visible on other nodes?
> 
> Yes, using ls. But no, using ls *.slc
> 
> RR> What "ls" are you using?  Is it built into your shell?
> 
> The Linux ls. I am not sure if it is built into my shell.
> I am not sure if I understand this question.

Some shells have a builtin "ls".  For example tcsh has a builtin ls-F that 
people sometimes alias ls to.  So if you run "/bin/ls *.slc" do you see 
this?  Just want to make sure that I understand what binary is doing this.

> RR> Could you send an strace of both the failing and successful commands?
> 
> Not sure what you wanted.

"strace" is a program that captures the system calls used by a program.  
You use it like "strace -o strace-ls.out ls *.slc".  That output would be 
useful to us.

Thanks,

Rob