← Back to team overview

openstack team mailing list archive

Re: [Swift] Storage Server Redirection



The blueprint of object-variation is able to allow an object to have
variant versions which are derived from the original master object. The
access to the master object can be smartly adapted or redirected to matched
version for different clients (web browser, mobile app,  video player). One
possible solution is that you can add system metadata for master object to
refer to those variant versions. When proxy server get client information
from HTTP headers, it will relay it to object server so that it can
determine which version is a better choice based on system metadata. If
current version is not good choice, reply a redirect response to tell proxy
server the right one. Another solution can be based on object versioning
feature to pick up the right version for client based on its request


-Edward Zhang

             "Luse, Paul E"                                            
             l.com>                                                     To
             Sent by:                  Peter Portante                  
             "Openstack"               <peter.a.portante@xxxxxxxxx>,   
             <openstack-bounce                                          cc
             s                         "openstack@xxxxxxxxxxxxxxxxxxx" 
             +zhuadl=cn.ibm.co         <openstack@xxxxxxxxxxxxxxxxxxx>
             m@lists.launchpad                                     Subject
             .net>                     Re: [Openstack] [Swift] Storage 
                                       Server Redirection              
             06/03/2013 08:41                                          

Hi Peter,

I'd ask that Edward may want to chime in on object variation as I may not
understand that one correctly - again I'm proposing to just implement the
redirect response blueprint and reference object variation as a potential
user simply because John mentioned it in his original write-up of the
redirect response blueprint.  Edward?

Wrt your second question, there are copies of the rings on object servers
as well and they are used for lookup in various scenarios, replication for
one.  As an example take a small 6 node cluster where all servers are
running all services, if an admin adds/deletes a device from the ring they
do so on one of the nodes and then copes the .gz file to all the others.
During the window when the first node detects the new ring file and .gz
files are copied to all the other nodes and detected/brought into memory,
any responses fielded by the proxy service on the other nodes will have
older ring info than the first node updated.  Another case could be a
larger cluster where there's a proxy tier and a capacity tier (not all
services running everywhere).  The admin could update the ring on any node
and copy the ring files on the nodes in any order - if any capacity node
gets an update before any proxy node you can run into the same scenario.
The ring versions blueprint was John's s!
 o he may have additions/corrections....


-----Original Message-----
From: Peter Portante [mailto:peter.a.portante@xxxxxxxxx]
Sent: Monday, June 03, 2013 3:50 AM
To: Luse, Paul E
Cc: openstack@xxxxxxxxxxxxxxxxxxx
Subject: Re: [Openstack] [Swift] Storage Server Redirection

Hi Paul,

Can you explain more about the two use cases referenced?

Object Variation seems like a nice idea, but can you clarify how the
mechanism would work to get at the variations? Any examples? From the
current descriptions, it seems like the LFS Patch could help here without
adding anything to the object server.

Regarding the ring updates, given that rings are used by the proxy servers,
and as far as I can tell, the object servers don't know about the rings,
how would object servers have more up-to-date knowledge to perform the



On Fri, May 31, 2013 at 7:53 PM, Luse, Paul E <paul.e.luse@xxxxxxxxx>
> I'm looking at tacking this item:
> https://blueprints.launchpad.net/swift/+spec/support-storage-server-re
> directs
> and wanted to get some feedback on the following observations/thoughts:
> 1) This is a capability that would be checked in independent of other
> blueprints that might use it (2 are mentioned in the link above) and
> unit test code would be the only way to initially exercise it; it
> essentially enables other activities at this point
> 2) The basic idea is that an object server (via middleware or
> otherwise) will be given the ability to respond to a request to
> indicate 'not me but I know who should handle this'.  I'm thinking
> this makes more sense as a 5xx response with additional information
> (partition, nodes) about the route included in the response body (as
> opposed to a 3xx code)
> 3) The proxy server will be modified to process the response
> accordingly but using the partition, nodes info from the response as
> opposed to
> object_ring.get_nodes() to determine which nodes to use
> 4) Protection will be required to avoid endless redirection loops
> 5) This applies only to GET operations
> Appreciate any thoughts/feedback.,  In addition to the two usages of
> this capability referenced in the blueprint I think there's applicable
> to another Tiering blueprint which interests me as well.
> Thanks
> Paul
> _______________________________________________
> Mailing list: https://launchpad.net/~openstack
> Post to     : openstack@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~openstack
> More help   : https://help.launchpad.net/ListHelp

Mailing list: https://launchpad.net/~openstack
Post to     : openstack@xxxxxxxxxxxxxxxxxxx
Unsubscribe : https://launchpad.net/~openstack
More help   : https://help.launchpad.net/ListHelp

GIF image

GIF image

GIF image

Follow ups