Part 2 of 6: blueprint general-bare-metal-provisioning-framework
In baremetal provisioning, one nova-compute manages multiple bare-metal
machines. A bare-metal machine does not run openstack at all.
Previously, bare-metal provisioning used text files to store information
of bare-metal machines. In this patch, a MySQL database is used to store
the information. We target only MySQL database. The DB is designed to
support PXE/non-PXE booting methods, heterogeneous hypervisor types, and
architectures. Using a MySQL database makes maintenance and upgrades
easier than using text files. The DB for bare-metal machines is
implemented as a separate DB from the main Nova DB. The DB can be on any
machines/places. The location of the DB and its server needs to be
specified as a flag in the nova.conf file (as in the case of glance).
There are a couple of reasons for this approach. First, the information
needed for bare-metal machines is different from that for non-bare-metal
machines. With a separate database for bare-metal machines, the database
can be customized without affecting the main Nova DB. Second, fault
tolerance can be embedded in nova-compute. Since one nova-compute
manages multiple bare-metal machines, fault tolerance of a nova-compute
node is very important. With a separate DB for bare-metal machines,
fault-tolerance can be achieved independently from the main Nova DB.
Replication of the bare-metal DB and implementation of fault-tolerance
are not part of this patch. The implementation models nova and its DB as
much as possible. The bare-metal driver must be upgraded to use this DB.
Change-Id: I7b7ba1903a672a50c567f95fc6554d119463b0c5
Co-authored-by: Mikyung Kang <mkkang@isi.edu>
Co-authored-by: David Kang <dkang@isi.edu>
Co-authored-by: Ken Igarashi <igarashik@nttdocomo.co.jp>
Co-authored-by: Arata Notsu <notsu@virtualtech.jp>
ee61bcb540