On Tue, Jun 8, 2010 at 1:10 PM, Matt Keenan <[email protected]> wrote: > Apologies, > > Forgot to add external link : > > http://hub.opensolaris.org/bin/download/Project+caiman/auto_install/ai-multi-design-0.1.pdf
For data pools, you say you don't support cache/log/spare devices. I see this as a significant omission and an inconsistency. I can see a reasonable limitation being that you only support 1- to N-way mirrored root pools; if you're going to support data pools at all then allowing cache/log/spare devices seems essential. (And not terribly difficult.) Besides, I expect such pool configurations to be very common. There's an example in 3.2.4.1: zpool create mypool mirror d1 d2 raidz d3 d4 d5 d6 seriously, I hope this would be rejected! There's also mirror c0t0d0 c0t1d0 c0t2d0 it may be a minor thing, but I immediately think of a root pool when given those devices, and that example then suggests that I can use EFI labelled disks rathe than SMI partitions. Appendix 1: default name is rpool - does this mean the name of the root pool can be changed? Is the default name of a data pool still rpool? Comment above "zpool_file_system_properties" - is that zpool create or zfs create? And -O or -o? (It's not immediately obvious why there are two sets of property settings here, when dataset properties can also be specified later on.) "Pool type, concatenated or mirror. Default if this element is specified is concatenation." Missing not? And the list is longer than just concatenated or mirror. Target devices - does this mean that if you just tell it to create a raidz, without listing devices, then it will just pick some at random? That seems horribly broken - I would expect that to just fail with an error. If you want auto-selection, then I would prefer to say so explicitly. Thanks, -- -Peter Tribble http://www.petertribble.co.uk/ - http://ptribble.blogspot.com/ _______________________________________________ caiman-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/caiman-discuss
