On Mon, Dec 02, 2013 at 12:23:02PM +0630, chandan wrote: > This patch implements the *core* of the idea suggested at > https://btrfs.wiki.kernel.org/index.php/Project_ideas#Linear_chunk_allocation_mode. > Other required changes (e.g. balance/restripe) will be made based on > the reviews obtained for this patch. > > On a multi-disk filesystem instance using single mode storage, we could > minimize the data loss (due to a hard drive failure) by filling up a disk > completely before starting to use another disk.
The usecase sounds ok, but I don't agree the incompat bit is the way to implement that, this is not a format change that would make it impossible to read the data on older kernels. I assume you want to make the allocation policy permanent and nothing else was available. We're in the process of getting the per-object properties ready and I think this will fit your needs. You can use a temporary mount option to set the allocator policy for now. There's a related project idea that would build upon the same infrastructure: https://btrfs.wiki.kernel.org/index.php/Project_ideas#Better_data_balancing_over_multiple_devices_for_raid1.2F10_.28allocation.29 > Chunk allocation is performed based on the increasing order of devid i.e. > the device with the lowest devid will be used for allocating the new chunk. This does not look very flexible and highly depends on the order in which the devices are added. I think a device property like 'allocation priority' would let the user chose the device order. david -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
