On Fri, Oct 7, 2016 at 5:26 AM, John Spray <jspray@xxxxxxxxxx> wrote: > On Thu, Oct 6, 2016 at 7:59 PM, Gregory Farnum <gfarnum@xxxxxxxxxx> wrote: >> On Wed, Oct 5, 2016 at 2:14 PM, Zack Cerza <zack@xxxxxxxxxx> wrote: >>> Hi all, >>> >>> You've probably heard that we've been working on replacing the >>> gitbuilders with a new build system. I recently did some work [0] in >>> teuthology to support that new system, and that work got merged >>> yesterday. Here's an overview of its current state: >>> >>> The way it works is that it talks to Shaman [1] to locate the package >>> repository to use. It is not enabled by default as it isn't yet able >>> to cover all gitbuilder use cases. It can be enabled by setting >>> "use_shaman: true" in either your teuthology.yaml or your job's yaml. >>> It can be set for an entire suite-run by doing something like: >>> >>> $ echo "use_shaman: true" > ./shaman.yaml >>> $ teuthology-suite -v -c jewel -s upgrade/jewel-x -m vps -k distro >>> $PWD/shaman.yaml --dry-run >>> >>> Of course, later on when we can fully replace gitbuilders this >>> behavior will be the default. At some point after that we'll be >>> decommissioning the gitbuilders. >>> >>> Please note that using it in the following scenarios will *not* do >>> what you want: >>> >>> * Any job that requires one of our testing kernels, as the new system >>> isn't building our kernels yet >>> * Any job that uses the install task for samba, for the same reason >>> * Any job that uses ceph-deploy, as ceph-deploy isn't using the new system yet >> >> I think Alfredo mentioned this in his thread, but we also don't seem >> to be building the "notcmalloc" flavors, so don't try and run any >> valgrind tests against it. >> Are there other flavors we know are missing? >> >> An FS run (http://pulpito.ceph.com/gregf-2016-10-05_20:40:25-fs-master---basic-mira/) >> had quite a few other failures, but just scanning the errors I don't >> think any of the others are because of the build system. > > I believe the java failures there are probably because they're not > getting built in the new system yet? I think there was a ticket but I > don't have it to hand. We are now (since last week) building libcephfs-java http://tracker.ceph.com/issues/17397 > > John > >> -Greg >> >>> >>> Please try to give this new system some exercise. It's been tested >>> well but of course we can't necessarily anticipate everything. >>> Teuthology bugs, as always, should go here [2] and bugs related to the >>> new build system should go here [3]. > _______________________________________________ > Sepia mailing list > Sepia@xxxxxxxxxxxxxx > http://lists.ceph.com/listinfo.cgi/sepia-ceph.com -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
