[Bugs] [Bug 1471309] New: non-atomic file move operation on same volume

bugzilla at redhat.com bugzilla at redhat.com
Sat Jul 15 00:14:31 UTC 2017


https://bugzilla.redhat.com/show_bug.cgi?id=1471309

            Bug ID: 1471309
           Summary: non-atomic file move operation on same volume
           Product: GlusterFS
           Version: mainline
         Component: core
          Severity: high
          Assignee: bugs at gluster.org
          Reporter: couture.danny at gmail.com
                CC: bugs at gluster.org



Created attachment 1298720
  --> https://bugzilla.redhat.com/attachment.cgi?id=1298720&action=edit
simple script trying to discover non-atomic move

Description of problem:
it is possible to see multiple intermediate state while moving a file over one
that already exists even if both the source and the destination are on the same
volume.

normally, on ext4, moving a file is an atomic operation where either the src
file is present with uncorrupted readable content or the destination is present
with uncorrupted readable content.

Version-Release number of selected component (if applicable):
3.10
mainline

How reproducible: 
100%

Steps to Reproduce:
1. run the attached script twice in parallel on the same folder of the same
gluster volume

Actual results:
multiple different results are currently possible

for a volume with a single brick, I get this error
md5sum: ./test: Operation not permitted
non atomic move!!! actual md5:  vs expected 0f343b0931126a20f133d67c2b018a3b

for a volume with multiple bricks using tcp transport, I get another error
md5sum: ./test: No such file or directory
non atomic move!!! actual md5:  vs expected 0f343b0931126a20f133d67c2b018a3b

Expected results:
same as ext4, no errors should ever happen, once created, the file should
always exist since its only overwritten... and its content should either be the
old content, or the new content, but not some intermediate state.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
You are the assignee for the bug.


More information about the Bugs mailing list