> > > It's used in stop_deployment provision stage [0] and for control reboot > [1]. > > > Is it a fall back mechanism if the mcollective fails? > > Yes it's like fall back mechanism, but it's used always [2]. >
As I remember it the use of SSH for stopping provisioning was because of our use of OS installers. While Anaconda was running the only access to the system was with SSH. > > > That might have been a side effect of cobbler and we should test if it's > > still an issue for IBP. > > As I can see from the code partition table always is wiped before > provision [3]. > There were some intermittent issues with provisioning failures (particularly with Ceph as I recall) when we only wiped the disks before provisioning. These failures were caused by stale LVM and RAID metadata. Doing it when the nodes were deleted and again before provisioning fixed these problems. I'm not sure this was a side-effect of our old provisioning method either. When we relied on the OS installers we generated kickstart and preseed files that just used the standard LVM and mdadm utilities to partition the system. Thanks, -Ryan
__________________________________________________________________________ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
