rgw: Multipart upload may failed when object size is large

[mdw-at-linuxbox]

[ revised to use common rgw logic to parse oids. - Marcus Watts <mwatts@redhat.com ]

Fixes:http://tracker.ceph.com/issues/21565

Signed-off-by: Sibei Gao gaosb@inspur.com

[mattbenjamin]

@gaosibei , @mdw-at-linuxbox confirms this fix to be important, but I'm a bit confused how it arises, and also about the string parsing as a mechanism; can you provide some detail on how / when this case reproduces, and maybe on the choice of approaches for solving?

[mattbenjamin]

@gaosibei ok, I understand the reproducer from the tracker; I think we need (and @yehudasa agrees we need) a different mechanism to detect such a case

[gaosibei]

@mattbenjamin

Do you need me to provide a new method to fix this bug？

[mattbenjamin]

@gaosibei yes, we'd like a new mechanism--if protocol change is required, we can motivate it in Mimic

[gaosibei]

@mattbenjamin Before geive a new method. I want to konw What conditions need to meet for the new mechanism? And what are the disadvantages of the current method?

[mattbenjamin]

@gaosibei my intuition was, we shouldn't need to parse object names in order to identify the full operation being performed is completing a multipart upload; @mdw-at-linuxbox , @yehudasa , do you want to amplify?

[gaosibei]

@mattbenjamin @mdw-at-linuxbox
Now, I give a new method to fix this bug which don't need parse object names.

The PR is: #17960

[liewegas]

closing this on preference to #17960