Carl and All,

SWMR can run only on the file system that guarantees order of operations (for 
example, NFS doesn’t guarantee it). HDF5 1.10.0 was never tested on zfs and, my 
guess is that operations ordering may be a problem.

Please see two documents (come with the HDF5 1.10.0 source or can be found in 
our repository)

https://svn.hdfgroup.org/hdf5/tags/hdf5-1_10_0/test/POSIX_Order_Write_Test_Report.pdf
https://svn.hdfgroup.org/hdf5/tags/hdf5-1_10_0/test/SWMR_POSIX_Order_UG.txt

for more details.

We also provide a “twriteorder" test that is built by the “make" command to 
check if ordering is a problem. Test’s success doesn’t guarantee that the 
system supports ordering, but the failure definitely indicates that SWMR will 
not work on the system. If someone on this FORUM will come up with the better 
test design and contribute the code, we would be more than happy to accept it.

We will separate SWMR tests from other C tests in 1.10.1 release and will try 
to detect if SWMR tests are run on the appropriate file system. For users that 
are not interested in the SWMR feature or do not have  the right file system, 
SWMR tests will not run automatically as they do now.

Thank you!

Elena


~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Elena Pourmal  The HDF Group  http://hdfgroup.org
1800 So. Oak St., Suite 203, Champaign IL 61820
217.531.6112
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~




On Apr 24, 2016, at 12:28 PM, Carl Ponder 
<[email protected]<mailto:[email protected]>> wrote:

On 04/24/2016 09:42 AM, Dana Robinson wrote:
The errors happen with the GCC, Intel & PGI compilers, using MVAPICH2 or 
OpenMPI, so if it's an issue with my software stack, it would have to be deeper 
than these.

    Creating skeleton data file for test...
    File created.
    1545: continue as the writer process
    dataset rank 3, dimensions 0 x 256 x 256
    1545: child process exited with non-zero code (1)
    Error(s) encountered
    HDF5-DIAG: Error detected in HDF5 (1.10.0) thread 0:
      #000: H5F.c line 579 in H5Fopen(): unable to open file
        major: File accessibilty
        minor: Unable to open file
      #001: H5Fint.c line 1208 in H5F_open(): unable to read superblock
        major: File accessibilty
        minor: Read failed
      #002: H5Fsuper.c line 443 in H5F__super_read(): truncated file: eof = 
526815, sblock->base_addr = 0, stored_eof = 33559007
        major: File accessibilty
        minor: File has been truncated
    H5Fopen failed
    read_uc_file encountered error
On 04/24/2016 09:42 AM, Dana Robinson wrote:
What file system are you testing on? Is it a network file system like NFS, AFS, 
or SMB?
That test was added in HDF5 1.10.0 and tests single-writer/multiple-readers 
(SWMR) functionality. Since that is a new feature for 1.10.0, the test is not a 
part of the HDF5 1.8 release.
Dana -- here's what I get from the "mount" command:
cmpool on /cm type zfs (rw,relatime,xattr,noacl)
I don't know how robust our filesystem/fileserver is.
I know we're not running out of space.
Thanks,

                    Carl Ponder

________________________________
This email message is for the sole use of the intended recipient(s) and may 
contain confidential information.  Any unauthorized review, use, disclosure or 
distribution is prohibited.  If you are not the intended recipient, please 
contact the sender by reply email and destroy all copies of the original 
message.
________________________________
_______________________________________________
Hdf-forum is for HDF software users discussion.
[email protected]<mailto:[email protected]>
http://lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
Twitter: https://twitter.com/hdf5

_______________________________________________
Hdf-forum is for HDF software users discussion.
[email protected]
http://lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
Twitter: https://twitter.com/hdf5

Reply via email to