On Mon, Jul 11, 2011 at 10:15:12PM +0200, Gabor Gombas spake thusly: > Since the AoE protocol does not support barriers, and AFAIK support for the > FLUSH ATA command was never implemented, the client kernel can do just one > thing: stop sending new commands, wait until all pending commands finish, and > really-really hope that the server did commit the data to disk, even if it > got no indication to do so.
Are there any plans to fix this? Is it even technically possible? It seems that Coraid would want to remove any doubt as to using AoE for "enterprise" use. This: http://lwn.net/Articles/283161/ has a good explanation of the history of write barriers in Linux. My understanding is that until recently RedHat had been turning off write barriers in the kernel anyway yet people still ran their journalling filesystems and databases etc. just fine. RHEL6 seems to have write barriers enabled for all filesystems that support them. A sync should ensure consistency as long as the backing disk system actually gets the data onto disk during the sync, right? My datacenter has had power issues lately so I am paying careful attention to this sort of thing. -- Tracy Reed Digital signature attached for your safety. Copilotco Professionally Managed PCI Compliant Secure Hosting 866-MY-COPILOT x101 http://copilotco.com
pgpQjwQ7eziti.pgp
Description: PGP signature
------------------------------------------------------------------------------ AppSumo Presents a FREE Video for the SourceForge Community by Eric Ries, the creator of the Lean Startup Methodology on "Lean Startup Secrets Revealed." This video shows you how to validate your ideas, optimize your ideas and identify your business strategy. http://p.sf.net/sfu/appsumosfdev2dev
_______________________________________________ Aoetools-discuss mailing list Aoetools-discuss@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/aoetools-discuss