Re: [PATCH 0/6] Reshape fixes (expansion)

[NeilBrown <neilb@suse.de>]

On Tue, 08 Feb 2011 14:30:28 +0100 Adam Kwolek <adam.kwolek@intel.com> wrote:

> The following series adds some fixes for reshape (expansion).
> 
> A while ago mdadm unit tests (suits 12 and 13) works correctly, now I've found problems.
> I've learned that containers created for raid5 and raid0 has different number of blocks in mdstat.
> I've put mdstat partial output to patch description:  UT FIX: imsm container can have different blocks number
> I'm not 100% sure that this is fix (for changed behavior), not workaround for container creation problem.
> This fix is first of fixes (3) that enables reshape (12, 13) UT.
> If you find this more as workaround, this patch is good point for discussion what change is correct.
> 
> Patches:
>       FIX: compare blocks on all data disks
>       FIX: Compare the same units
> makes changes for blocks computing verification. After those 2 more  patches UT (12, 13) works. 
> Please let me know what you are thinking about them. It is possible that this is not final patch also,
> but I've tried to find out what this condition should do, and change code to archive it (IMHO).
> 
> Rest of patches addresses other issues, I can observe during expansion.
> 
> BR
> Adam
> 
> ---
> 
> Adam Kwolek (6):
>       FIX: md runs recovery instead reshape for growing single disk raid0 array
Applied.

>       FIX: Container can be left frozen
Applied.

>       FIX: compare blocks on all data disks
>       FIX: Compare the same units
No, I don't like these.
The point here is to compare the amount of data to be backed up with the
size of one of the spare devices, because in some cases the data will be
backed up to a spare device.
If the size of the data to be backed up is bigger than half the size of a
spare device, then it doesn't really fit and something is wrong.
Normally the amount of data to be backed up is much much smaller than the
size of any device.

Maybe we just need to make the test use larger devices or smaller chunk size,
or similar.


>       UT FIX: imsm container can have different blocks number
Applied (the block numbers here are completely meaningless, so ignoring them
is good).
>       imsm: FIX: Wrong output string format
Applied.

All pushed out to devel-3.2

Thanks.

NeilBrown