Hi Jean,
Jean McCormack wrote: > Jean McCormack wrote: >> jan damborsky wrote: >> >>> Hi Jean, >>> >>> the changes look good to me. >>> >>> Uut of curiosity, since DIST_PROTO is created on zfs dataset, >>> might be 'used' property utilized for obtaining size information ? >>> Something like >>> # zfs get -Hp used <proto_zfs_dataset> >>> >>> I am not asking to change the code, just curious :-) >>> >> I think I could use that. I'm curious though. If I do >> df -k /rpool/dc-builds >> >> Filesystem kbytes used avail capacity Mounted on >> rpool/dc-builds 63479808 2102448 54479705 4% >> /rpool/dc-builds >> >> >> yet zfs get -Hp used rpool/dc-builds yields: >> rpool/dc-builds used 2814928384 - >> >> >> Why are they different? Even compensating for the df -k being kbytes >> it's not the same number. >> (2102448 x 1024 = 2152906752) >> >> My guess is that there are subtle differences between what they are >> giving the size of. I'm >> just not sure what that subtlety is. >> > The difference is the size of the underlying filesystem for > bootcd_microroot. It's small at > this point in the build, about 26.5 K, but it's there. Just to be > clear, I have my build area > set up so that $DIST_PROTO = rpool/dc-builds is one filesystem and > $MICROROOT= > rpool/dc-builds/bootcd_microroot is a different one. Possibly that's > unconventional? > > So I'm not sure we do want to go with zfs get. Depends on whether my > setup is reasonable > or not. What do you think? I am sorry I didn't realize that there is additional filesystem defined in 'proto' area which should be excluded. My feeling is that $MICROROOT is not part of 'proto' area but I don't have reasonable arguments for this - I don't think it is unconventional. Since zfs 'used' property accounts for all descendents defined within dataset, it seems to me we can't take advantage of it in this case. Thank you, Jan
