GNOME Bugzilla – Bug 695731
ostree.gnome.org's testing code needs to be deployed on a physical server
Last modified: 2013-05-28 12:31:43 UTC
Right now on my laptop, booting a gnome-ostree VM in qemu takes about 5 seconds. On the server, due to nested virtualization plus ambient I/O activity, it takes 5 **minutes**. This really hampers my ability to do automated testing. I hear there's possibly new hardware coming...if there's anything we can do with partners/donors to accelerate that, I can dramatically increase the speed at which testing happens. Possibly we could split up the testing from the building - even a portion of a basic server with 4G of RAM and 100GB of disk would be enough to run qemu, as long as the host is KVM-capable.
Managing the new server: * Puppet sounds good for maintaining the package set and config - Package set: See https://bugzilla.gnome.org/show_bug.cgi?id=669772 for current package set. High level: need linux-user-chroot from EPEL, a big pile of -devel packages, and libguestfs. * Current "buildmaster" LDAP group should have ssh access * Add a "ostree" user, code gets deployed and run as that user * Deploy storage as /srv/ostree ? Current is 250GB, 500GB would be plenty for near future. * buildmaster users can sudo to ostree * Will also need an /etc/init.d/gnome-ostree init script so everything gets started on bootup (Colin to write this) * Also set up httpd exporting /srv/ostree; need SSL too
The physical machine for the new ostree service has been installed and configured already matching Colin's needs.
Yep, It's working very well, thanks!