Object remains on handoff after DELETE

Bug #1314799 reported by Chuck Thier
28
This bug affects 6 people
Affects Status Importance Assigned to Milestone
OpenStack Object Storage (swift)
New
Medium
Thiago da Silva

Bug Description

When a user issues a DELETE, and an object is on a handoff node, the handoff object does not get deleted. This normally wouldn't be a big issues, except a HEAD or GET on the object will still find it, thus not appearing to be deleted by the user, at least until replication gets to the object. I'm not advocating that we issue a DELETE to all handoff nodes as well, but should figure out some way to handle this case.

It could be argued to chalk it up to eventual consistency, but I think it warrants trying to handle in a reasonable way.

Revision history for this message
John Dickinson (notmyname) wrote :

I wonder if we can differentiate at the proxy server an object server saying "404 I didn't find it" and "404 I found a tombstone with timestamp X". Then the proxy can do the right thing (ie probably keep looking into handoffs and only return if a newer non-404 isn't returned).

Revision history for this message
Chuck Thier (cthier) wrote :

Yeah that was one possibility that we were discussing here as well

Revision history for this message
Samuel Merritt (torgomatic) wrote :

Pretty sure the object server 404 response has an X-Timestamp if there was a tombstone, and no X-Timestamp if there wasn't.

Revision history for this message
Thiago da Silva (thiagodasilva) wrote :

I think this bug is a duplicate of bug #1560574. The behavior is explained in a different way, but I believe the solution will turn to be the same.

Changed in swift:
importance: Undecided → Medium
assignee: nobody → Thiago da Silva (thiagodasilva)
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.