>>Thanks for the update!
>>
>>How will we handle repositories like RPMFusion, EPEL, ELREPO and REMI?
>>
>>// Roger

I plan on using separate labels for each. There will be a separate mirrorball 
config and job for each import.

--Brett

31 juli 2014, Brett Smith <[email protected]> skrev:
> Build Nodes --
>
>
> New Build Nodes are configured and active. Old virtual build nodes are out of 
> the cluster
>
>
> build53 [is: x86 x86_64] (36 slots)
>
> build54 [is: x86 x86_64] (36 slots)
>
> build55 [is: x86 x86_64] (36 slots)
>
> build56 [is: x86 x86_64] (36 slots)
>
>
> Sync of repos --
>
>
> Wrote a simple rsync scripts for Fedora and CentOS and dropped them in 
> /home/mirror/bin so we can cron the syncs.
>
>
>
> Fedora 20 --
>
>
> Processing of Fedora continues. It was very slow going until this week when I 
> took time to make mirrorball a little more efficient.
>
>
> There have been a few checksum errors with rpms but other than that chugging 
> along with processing Fedora 20.
>
>
> CentOS 7 --
>
>
> Started syncing the CentOS 7 beta today. The plan is to use Fedora 20 to 
> import the beta once we finish processing Fedora 20. The beta will go on a 
> beta label and hopefully I will have it working well enough to start the 
> CentOS 7 import once it is available.
>
>
> Plan on importing centosplus, extras, and fasttrack to separate labels.
>
>
> Mirrorball updates --
>
>
> Mirrorball is the python code we use to update manifests for encapsulated 
> distros. It looks for packages to update by comparing the yum repo to the 
> conary repo. If a pkg is newer in the yum repo then on the conary label it 
> updates the package manifest, commits, and sends it to the list of packages 
> to be built.
>
>
> I am pretty sure I properly fixed mirrorball's long pause every time it 
> started on a new chunk of updates
>
> To be clear the pause in mirrorball was silliness on my part. I break the 
> update up into 96 packages at a time... this allows me to load the 4 build 
> nodes evenly with a build per real processor. Every time mirrorball swung 
> around to do another 96 it makes a Conary Source to rpm Source map... it was 
> taking a long time (like 30 minutes). Turns out when you instantiate the 
> conaryhelper to do this it clears the cache so it was rebuilding the 
> sourceMaps from the label every 96 packages... it was never a problem until 
> Fedora and its 65000 pkgs.
>
>
> No need for that during a single update run so now we do it at the beginning 
> of the update and pass the sourceMaps along until all the updates are 
> processed then if we have to do anything like order or build a group after 
> the update the cache gets cleared and rebuilt (hopefully with the latest 
> troves we built)
>
>
> --
>
> Brett C. Smith
> <[email protected]>
> Sr Software Developer
> Platform Deployment Technologies
> (919)531-6635 -- x16635
> _______________________________________________
> Foresight-devel mailing list
> [email protected]
> <https://lists.foresightlinux.org/mailman/listinfo/foresight-devel>
>
_______________________________________________
Foresight-devel mailing list
[email protected]
https://lists.foresightlinux.org/mailman/listinfo/foresight-devel
_______________________________________________
Foresight-devel mailing list
[email protected]
https://lists.foresightlinux.org/mailman/listinfo/foresight-devel

Reply via email to