[Bugs] [Bug 1141379] Geo-Replication - Fails to handle file renaming correctly between master and slave

[bugzilla at redhat.com]

https://bugzilla.redhat.com/show_bug.cgi?id=1141379



--- Comment #10 from James Payne <jamoflaw at gmail.com> ---
(In reply to Aravinda VK from comment #9)
> (In reply to James Payne from comment #7)
> > (In reply to Aravinda VK from comment #5)
> > > Please try the 3.6 beta 3 version. Two patches landed related to the rename
> > > and retry issues.
> > > 
> > > http://review.gluster.org/#/c/8879/
> > > http://review.gluster.org/#/c/8880/
> > 
> > Can confirm the issue still exists in beta3. 
> > 
> > v easy to replicate 
> > 
> > touch file && mv file file2
> > 
> > Will create 2 files on the geo-replica and only one on the primary cluster
> 
> With the above patches, GFID mismatch and rsync retry issues are resolved,
> Do you see any data loss in Slave? both file and file2 present in slave, Is
> data intact in file2?
> 
> Geo-rep has a known issue with Rename handling when renamed file fall into
> separate brick other than the brick in which it is created.
> 
> Summarized RENAME issues here,
> http://supercolony.gluster.org/pipermail/gluster-devel/2014-September/042316.
> html
> 
> Development is in progress to fix all these RENAME issues. Will update the
> bug with progress.



You can get data loss on slave in the following scenario:

touch file1 && mv file2

** Wait for replication

echo contents > file1


You will find on the geo replica that file1 is now no longer consistent with
the local copies. The same goes for change of attributes as well. I believe the
contents of file2 in the case above will be correct but file1 will never be
consistent with the local copy.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
Unsubscribe from this bug https://bugzilla.redhat.com/token.cgi?t=SFQvJpWydS&a=cc_unsubscribe