On Mon, Dec 3, 2012 at 10:25 AM, John Burwell <jburw...@basho.com> wrote: > Chiradeep, > > Please see my responses in-line below. > > Thanks, > -John > > On Nov 20, 2012, at 9:33 PM, Chiradeep Vittal <chiradeep.vit...@citrix.com> > wrote: > >> I must give kudos on the document >> (https://cwiki.apache.org/confluence/display/CLOUDSTACK/S3-backed+Secondary >> +Storage ). Quite complete >> >> If the intent is to make immutable assets (templates etc) visible across >> all zones: >> 1. Does the template_zone_ref table still make sense? Is it automatically >> updated? > > Yes, the template_zone_ref table is still required. Templates and ISOs must > be attached to a zone. From my understanding of the system design, removing > this requirement would require a fundamental data model/architectural change. > Therefore, cross zone templates are supported by creating a row in the > template_zone_ref table associating the template to the zone. These > propigation operations are performed upon template and zone creation. This > behavior pre-dates this patch. This patch automates the operation of > propagating the data across zones. > > I have updated the data model section of the design document to describe the > data flow and role of each table. > >> 2. If I request a vm deployment from a template in a zone that has not yet >> sync'ed, what is the behavior? > > The short answer is that S3 download behaves similarly to Swift downloads. > Templates and ISOs are downloaded from S3 to NFS on-demand. When a template > is not found in the NFS volume, the system will attempt to downloaded it from > S3. As noted in the design download, this on-demand behavior may create a > slight lag when a template is initially accessed in a zone dependent on the > available bandwidth between the SSVM and the S3-compatible store. > >> 3. If I add a zone is the sync automatic to the new zone > > Yes, per the template propagation and on-demand download behaviors described > above, templates and ISOs will automatically be available in a newly created > zones. > > I explained this process in the Architecture/Design section, and have updated > it based on on my response here. Please let me know if it is needs further > clarification. > >> >> >> >> >> On 11/18/12 10:51 PM, "John Burwell" <jburw...@basho.com> wrote: >> >>> All, >>> >>> I have submitted the initial revision of a patch to support S3 backing >>> NFS secondary storage (https://reviews.apache.org/r/8123/). I have >>> opened ticket CLOUDSTACK-509 to track the testing and delivery of the >>> enhancement. I have also created a design document in the wiki roughly >>> following design document template. Since I am not a committer, it did >>> not seem appropriate for me to add a reference from the roadmap to the >>> ticket. >>> >>> In its current state, I have tested the single zone template and ISO >>> functionality on devcloud. I am working to configure a multi-zone >>> integration testing environment to complete all test scenarios outlined >>> in the design document. My plan is to refine the arch per community >>> patch concurrently with the completion of integration testing and bug >>> squashing. >>> >>> I look forward to your feedback, >>> -John >> > >
Edison, Chiradeep, Rohit - Can you guys spend some time on the latest proposed patch [1] to see if the outstanding issues / questions have been resolved? If it looks good, we should get the changes committed to the master branch. I'd like to get this feature into the 4.1.0-incubating release. -chip [1] https://reviews.apache.org/r/8123/
