excessive MC downloads bug

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

excessive MC downloads bug

Eliot Miranda-2
Hi All,

   more whitewash... I have a theory about the bug whereby one...

1. commits a package to a remote repository
2. opens the repository, browses the package and finds that MC downloads the package again even though its just been written to package-cache.


Is it that when the package is uploaded to the remote repository it is generated again and hence has a different UUID from the one in package-cache?  Hence when looking for a version the system rejects the one in package-cache because it has a different UUID?

If this is so the fix is simple; when copying a package from one repository to another, the *existing file* should be copied, not regenerated.  That's got to be better right?  Less work all round.
--
best,
Eliot


Reply | Threaded
Open this post in threaded view
|

Re: excessive MC downloads bug

Chris Muller-3

On Fri, Jul 18, 2014 at 2:27 PM, Eliot Miranda <[hidden email]> wrote:
Hi All,

   more whitewash... I have a theory about the bug whereby one...

1. commits a package to a remote repository
2. opens the repository, browses the package and finds that MC downloads the package again even though its just been written to package-cache.


Is it that when the package is uploaded to the remote repository it is generated again and hence has a different UUID from the one in package-cache?  Hence when looking for a version the system rejects the one in package-cache because it has a different UUID?

No.  The UUID's are the only glue it has to BIND MC's distributed model together.  That's the only way we know one version is the same as another in a different repository.
 
If this is so the fix is simple; when copying a package from one repository to another, the *existing file* should be copied, not regenerated.  That's got to be better right?  Less work all round.

It's already a copy, but I know the problem you're talking about; its very annoying and creates unnecessary hits on our servers.  My theory is it's hitting "MCRepositoryGroup default" under certain conditions to look for the repository version.  As you know, "RepositoryGroup default" is the list you see in the MC browser when no package is selected.  The Group object acts like a repository itself by enumerating through all of its repositories, looking for the first one that can satisfy a request..