On May 27, 2011, at 8:56 AM, Christopher Brooks wrote:

> Hi Ruediger,
> 
>>> 1. With the addition of a work around such as setting the database
>>> sequence counter manually, do you still see this as a release
>>> blocker?
>> No. This would be a workaround. But it is only a workaround and we 
>> should look for a better solution very soon.
> 
> Great, I agree on both accounts.  Ideally we should go back to using
> uuids when sequence isn't important, and I think in MPs a sequence
> identifier isn't important.
> 
> I'll leave this for someone who understands that part of the codebase
> better than I to deal with though :)
> 
> As per our meeting yesterday, I deprioritized this task to
> critical and added a second fix version of 1.3.
> 
> Chris


I thought mediapackages already have UUIDs rather than sequenced IDs.  If this 
isn't the case, please file a bug.  Workflow instances use sequences, since 
there's no intent to federate them.

If we want to provide a relatively painless upgrade path for 1.3, the 1.2 data 
storage must be upgradable.  I did a quick analysis of what's missing in the 
trunk, and I *think* search was the last problematic service.  So I filed and 
fixed MH-7791, which enables removal and regeneration of the search index.

So for 1.3, we may be able to simply run a DDL/DML script, remove the various 
solr indexes, deploy the new code, and start Matterhorn.  In the absence of a 
more exhaustive analysis and proposal, I think this is the best we can do for 
the (already delayed) 1.2 release.

Josh
_______________________________________________
Matterhorn mailing list
[email protected]
http://lists.opencastproject.org/mailman/listinfo/matterhorn


To unsubscribe please email
[email protected]
_______________________________________________

Reply via email to