Hi Alexander,
Please see a few of my comments inline.
Alexander Eremin wrote:
> On Fri, 2009-10-02 at 12:04 +0000, Alexander Eremin wrote:
>
>> Hi all,
>> Just played with Bug 6361 - Calculation of Sparc boot archive doesn't
>> count with dcfs compression.
>>
I just updated the bug with my understanding of the cause of the
problem.
>> If change overhead for sparc in bootroot_archive.py to ~1.062 I've got
>> a 114mb boot_archive with 23mb free space.
>> That's minimum, in other case I've got cpio errors in br-arch step
>>
Please note that the code in root_archive.py is for calculating the size
dynamically based on the bootroot content specified in the input manifest.
That need to work for all image types.
>> If I take usual root_archive script for DC /bootroot directory and add
>> to it line:
>> size=$(($size +
>> 20480))
>>
>>
>> before
>>
>>
>> /usr/sbin/mkfile ${size}k "$TMR"
>> ...
>>
>> I've got a 90mb boot_archive with 13mb free space.
>> Looks like DC python scripts produces bigger size for boot_archive
>> than usual root_archive cmd.
>>
>>
> On my sparc for bootroot directory ('du' there give me 109mb but I know
> that on zfs the disk usage is less than the sum of the file sizes)
> bootroot_archive.py calculate boot_root_size as 109mb before
> ti_create_target, tunefs and compression,
> root_archive - 68mb before mkfile, lofiadm,newfs and compression
> processes.
> root_archive give me smaller miniroot.
>
du doesn't return accurate size on ZFS.
http://opensolaris.org/os/community/zfs/faq/#whydusize
Thanks,
--Karen
