-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi, This is related to the recent thread about best practices in using shared storage, but coming at it from a slightly different angle. I''m setting up a pre-production/test environment using XCP; with how we plan on operating this system, there''s going to be some pretty rampant snapshots and cloning of some reasonably large VHDs. As such, I want to use file-based VHDs rather than LV-based, in order to take advantage of thin-provisioning to minimise disk space. I''m happy with the performance hit this causes. Further, I want to use shared storage so that I can have multiple hosts and can easily expand processing capacity as we spin up various instances, and do migrations. However, I''m not using shared storage for auto failover or hot spare type functionality; migration will be manually managed as required. So, from what I''ve been reading, I think I need one of the following two options: 1) NFS. Simple, understood technology. Low overhead, and the XAPI toolstack takes care of "sharing" the VHDs. 2) iSCSI, GFS(2), cLVM. Storage LUN(s) presented by iSCSI, turned into an LV using cLVM, formatted with GFS or GFS2, and this filesystem added as a "file" type SR. More complicated than NFS, and I''ve read there were some problems with GFS in this sort of scenario, to do with mounting via the loopback device. But that was back a few years, and may have been solved, either in GFS or in GFS2. Have I missed any other options? Just pointers in the right direction (keywords) is enough if that''s all you''ve got time for. Is there anything glaringly wrong with my briefly written understanding of the options? And does anyone have any comments on which is likely to be better? Thanks, - -- Craig Miskell Senior Systems Administrator Opus International Consultants Phone: +64 4 471 7209 "Opportunity is missed by most people because it is dressed in overalls and looks like work." -- Thomas A. Edison -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.9 (GNU/Linux) iEYEARECAAYFAky7vocACgkQmDveRtxWqnb26ACgtPdgxK2+3KKmf3SZUuSjm45B f64AnRpD23tfBYkPzrHW+JqXMohx8eZv =VyVX -----END PGP SIGNATURE----- _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Monday 18 October 2010 05:27:03 Craig Miskell wrote:> Hi, > This is related to the recent thread about best practices in using > shared storage, but coming at it from a slightly different angle. > > I''m setting up a pre-production/test environment using XCP; with how we > plan on operating this system, there''s going to be some pretty rampant > snapshots and cloning of some reasonably large VHDs. As such, I want to > use file-based VHDs rather than LV-based, in order to take advantage of > thin-provisioning to minimise disk space. I''m happy with the performance > hit this causes. > > Further, I want to use shared storage so that I can have multiple hosts and > can easily expand processing capacity as we spin up various instances, and > do migrations. However, I''m not using shared storage for auto failover or > hot spare type functionality; migration will be manually managed as > required. > > So, from what I''ve been reading, I think I need one of the following two > options: > > 1) NFS. Simple, understood technology. Low overhead, and the XAPI > toolstack takes care of "sharing" the VHDs. > > 2) iSCSI, GFS(2), cLVM. Storage LUN(s) presented by iSCSI, turned into an > LV using cLVM, formatted with GFS or GFS2, and this filesystem added as a > "file" type SR. More complicated than NFS, and I''ve read there were some > problems with GFS in this sort of scenario, to do with mounting via the > loopback device. But that was back a few years, and may have been solved, > either in GFS or in GFS2. > > Have I missed any other options? Just pointers in the right direction > (keywords) is enough if that''s all you''ve got time for. > > Is there anything glaringly wrong with my briefly written understanding of > the options? > > And does anyone have any comments on which is likely to be better? > > Thanks,AoE, http://en.wikipedia.org/wiki/ATA_over_Ethernet I''ve heard it''s very fast, but rather unpopular. Bart _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Monday 18 October 2010 05:27:03 Craig Miskell wrote:> Hi, > This is related to the recent thread about best practices in using > shared storage, but coming at it from a slightly different angle. > > I''m setting up a pre-production/test environment using XCP; with how we > plan on operating this system, there''s going to be some pretty rampant > snapshots and cloning of some reasonably large VHDs. As such, I want to > use file-based VHDs rather than LV-based, in order to take advantage of > thin-provisioning to minimise disk space. I''m happy with the performance > hit this causes. > > Further, I want to use shared storage so that I can have multiple hosts and > can easily expand processing capacity as we spin up various instances, and > do migrations. However, I''m not using shared storage for auto failover or > hot spare type functionality; migration will be manually managed as > required. > > So, from what I''ve been reading, I think I need one of the following two > options: > > 1) NFS. Simple, understood technology. Low overhead, and the XAPI > toolstack takes care of "sharing" the VHDs. > > 2) iSCSI, GFS(2), cLVM. Storage LUN(s) presented by iSCSI, turned into an > LV using cLVM, formatted with GFS or GFS2, and this filesystem added as a > "file" type SR. More complicated than NFS, and I''ve read there were some > problems with GFS in this sort of scenario, to do with mounting via the > loopback device. But that was back a few years, and may have been solved, > either in GFS or in GFS2. > > Have I missed any other options? Just pointers in the right direction > (keywords) is enough if that''s all you''ve got time for. > > Is there anything glaringly wrong with my briefly written understanding of > the options? > > And does anyone have any comments on which is likely to be better? > > Thanks,Ow yes, I use iSCSI with nothing on it, as I use shared block devices, not image files. It rules out the complexity of GFS, cLVM or OCFS2. You do need a clustering software to prevent guest booting from the same storage twice. Good luck, B. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users