GlusterFS as a Storage Backend for oVirt/VDSM
GlusterFS is a user-space distributed file system, capable of scaling to several peta-bytes. It aggregates various storage bricks over TCP/IP (and other supported transport layers) and interconnect into one large parallel network file system. Storage bricks can be made of any commodity hardware, for eg: x86-64 server(s).
oVirt platform is a richly featured virtualization management solution providing fully integrated management across virtual machines. It aims to provides large scale, centralized management for server and desktop virtualization.
oVirt uses VDSM as the node virtualization management API. VDSM acts as the oVirt's host agent. VDSM abstracts low level details of the underlying linux hosts and exposes a high level API to oVirt for managing the linux host.
In recent times, there have been efforts to integrate Gluster as a block backend
to QEMU. There were patches posted in qemu-devel to support gluster as a block backend for qemu.
For QEMU-Gluster integration to be consumable from oVirt, we needed to make VDSM understand Gluster as a storage domain/backend.
This talk will focus on what it takes to integrate GlusterFS as a storage domain
in VDSM, different ways of doing it and how VDSM can be made to exploit the QEMU-Gluster block backend support.
Draft Flow of the presentation (Total time = 45min)
1) High level overview of oVirt/VDSM - 5min
2) VDSM Storage Concepts ( Domain, pool, SPM etc) - 5min
3) Different Approaches to integrating Gluster with VDSM as a Storage Domain
3a) Sharedfs/PosixFs approach - 10min
3b) Using VDSM Hooks - 10min
3c) More Native approaches ( under discussion with community as of now) - 10min
4) Q&A - 5min