up2date, mirror repositories, and performance

seth vidal skvidal at phy.duke.edu
Thu Feb 19 05:15:08 UTC 2004


> Seth, I believe you are right on target here.
> 
> The only other thing I can think of is some kind of central augmentation of 
> the mirror list which would reflect the sync status as well as somewhat rank 
> the servers in terms of capabilities (a big server attached to an OC48 can 
> take a lot bigger load than a small server on a T1).  This sounds good but I 
> am not sure it is do-able.

well, you know. it's not really about a mirror being in sync it's more
about a repository being in sync. As a mirror can contain many
repositories in various stages of 'sync-ness'

so if the metadata format we've worked on can become a standard for
fedora core you could use the repomd.xml file as a way to check mirrors.
It has timestamps and md5sums in it. The mirror master or some client
could download this <1K file a set of mirrors compare it to the one on
the mirror master and know (within a fair margin of error) which one's
were in sync.

With regard to your ranking question - ranking is best left to the
individual clients. Rank depends largely on how far and over what
networks you are distanced from the mirror.

-sv






More information about the fedora-test-list mailing list