[Ovirt-devel] image minimizer in ovirt-node-image kickstart

[Jim Meyering]

"Perry N. Myers" <pmyers at redhat.com> wrote:
> Noticed that there is a new section in the common-post.ks in
> ovirt-node-image using something call Image Minimizer.  I assume this
> is the whitelist/blacklist solution that the AOS team has been working
> on?
>
> Some of the output from that is suspect...  Here's what I'm seeing:
>
>> Running image-minimizer...
>> Unknown Command: touch
>> e2fsck 1.41.0 (10-Jul-2008)
>> Pass 1: Checking inodes, blocks, and sizes
>> Pass 2: Checking directory structure
>> Pass 3: Checking directory connectivity
>> Pass 4: Checking reference counts
>> Pass 5: Checking group summary information
>> ovirt-node-image: 3724/35200 files (1.1% non-contiguous), 52460/140800 blocks
>> e2fsck 1.41.0 (10-Jul-2008)
>> Pass 1: Checking inodes, blocks, and sizes
>> Pass 2: Checking directory structure
>> Pass 3: Checking directory connectivity
>> Pass 4: Checking reference counts
>> Pass 5: Checking group summary information
>> Directories count wrong for group #0 (441, counted=241).
>> Fix? yes
>>
>>
>> ovirt-node-image: ***** FILE SYSTEM WAS MODIFIED *****
>> ovirt-node-image: 3724/14080 files (2.1% non-contiguous), 51098/51098 blocks

I've seen that fsck-detected failure regularly, in the last couple days.
I presume you're building on a system with at least 2 cores, right?
I seem to recall that one fs-related phase was parallelized, so took
advantage of all cores on a multi-core system, so maybe there's a race.

> Don't know if the resulting fsck is caused by the image minimizer or
> not, but if someone else can replicate what I'm seeing here we need to
> look into this.
>
> Also, the "Unknown command: touch" seems to be coming from the image
> minimizer as well.

I proposed a work-around, then Alan proposed a better patch for that,
but it was not possible to ACK with "works for me" yesterday,
due to the early-boot problem you've just fixed.