On January 17, 2020 6:20:09 PM GMT+02:00, Nir Soffer <nsof...@redhat.com> wrote:
>On Fri, Jan 17, 2020 at 6:41 AM Strahil Nikolov <hunter86...@yahoo.com>
>wrote:
>
>> On January 17, 2020 12:10:56 AM GMT+02:00, Chris Adams
><c...@cmadams.net>
>> wrote:
>> >Once upon a time, Nir Soffer <nsof...@redhat.com> said:
>> >> On Tue, Jan 7, 2020 at 4:02 PM Chris Adams <c...@cmadams.net>
>wrote:
>> >> > Once upon a time, m.skrzetu...@gmail.com
><m.skrzetu...@gmail.com>
>> >said:
>> >> > > I'd give up on the ISO domain. I started like you and then
>read
>> >the docs
>> >> > which said that ISO domain is deprecated.
>> >> > > I'd upload all files to a data domain.
>> >> >
>> >> > Note that that only works if your data domain is NFS... iSCSI
>data
>> >> > domains will let you upload ISOs, but connecting them to a VM
>> >fails.
>> >>
>> >> ISO on iSCSI/FC domains works fine for starting a VM from ISO,
>which
>> >is the
>> >> main use case.
>> >
>> >Okay - it didn't the last time I tried it (I just got errors). 
>Thanks.
>>
>> I have opened and RFE for ISO checksumming, as currently the uploader
>can
>> silently corrupt your DVD.
>>
>
>Can you share the bug number?
>
>
>> With gluster, I have an option to check the ISO checksum and
>> verify/replace the file, but with Block-based storage that will be
>quite
>> difficult.
>>
>
>Checksumming is a general feature not related to ISO uploads. But
>checksumming tools do not understand
>sparseness so you should really use a tool designed for compare disk
>images, like "qemu-img compare".
>
>Here is an example:
>
>1. Create fedora 30 image for testing:
>
>$ virt-builder fedora-30 -o fedora-30.raw
>...
>$ qemu-img info fedora-30.raw
>image: fedora-30.raw
>file format: raw
>virtual size: 6 GiB (6442450944 bytes)
>disk size: 1.15 GiB
>
>2. Create a checksum of the image
>
>$ time shasum fedora-30.raw
>991c2efee723e04b7d41d75f70d19bade02b400d  fedora-30.raw
>
>real 0m14.641s
>user 0m12.653s
>sys 0m1.749s
>
>3. Create compressed qcow2 image with same content
>
>$ qemu-img convert -f raw -O qcow2 -c fedora-30.raw fedora-30.qcow2
>...
>$ qemu-img info fedora-30.qcow2
>image: fedora-30.qcow2
>file format: qcow2
>virtual size: 6 GiB (6442450944 bytes)
>disk size: 490 MiB
>cluster_size: 65536
>Format specific information:
>    compat: 1.1
>    lazy refcounts: false
>    refcount bits: 16
>    corrupt: false
>
>This is typical file format used for publishing disk images. The
>contents
>of this
>image are the same as the raw version from the guest point of view.
>
>3. Compare image content
>
>$ time qemu-img compare fedora-30.raw fedora-30.qcow2
>Images are identical.
>
>real 0m4.680s
>user 0m4.273s
>sys 0m0.553s
>
>3 times faster to compare 2 images with different format compared with
>creating
>a checksum of single image.
>
>Now lets see how we can use this to verify uploads.
>
>4. Upload the qcow2 compressed image to a new raw disk (requires ovirt
>4.4
>alpah3):
>
>$ python3 upload_disk.py --engine-url https://engine/ --username
>admin@internal --password-file password \
> --cafile ca.pem --sd-name nfs1-export2 --disk-format raw --disk-sparse
>fedora-30.qcow2
>
>5. Download image to raw format:
>
>$ python3 download_disk.py --engine-url https://engine/ --username
>admin@internal --password-file password \
>--cafile ca.pem --format raw f40023a5-ddc4-4fcf-b8e2-af742f372104
>fedora-30.download.raw
>
>6. Comparing original and downloaded images
>
>$ qemu-img compare fedora-30.qcow2 fedora-30.download.raw
>Images are identical.
>
>Back to the topic of ISO uploads to block storage. Block volumes in
>oVirt
>are always aligned to
>128 MiB, so when you upload an image which is not aligned to 128 MiB,
>oVirt
>creates a bigger
>block device. The contents of the device after the image content are
>not
>defined, unless you
>zero this area during upload. The current upload_disk.py example does
>not
>zero the end of
>the device since the guest do not care about it, but this makes
>verifying
>uploads harder.
>
>The best way to handle this issue is to truncate the ISO image up to
>the
>next multiple of 128 MiB
>before uploading it:
>
>$ ls -l Fedora-Server-dvd-x86_64-30-1.2.iso
>-rw-rw-r--. 1 nsoffer nsoffer 3177185280 Nov  8 23:09
>Fedora-Server-dvd-x86_64-30-1.2.iso
>
>$ python3 -c 'n = 3177185280 + 128 * 1024**2 - 1; print(n - (n % (128 *
>1024**2)))'
>3221225472
>
>$ truncate -s 3221225472 Fedora-Server-dvd-x86_64-30-1.2.iso
>
>The contents of the iso image is the same as it will be on the block
>device
>after the upload, and
>uploading this image will zero the end of the device.
>
>If we upload this image, we can check the upload using qemu img
>compare.
>
>$ python3 upload_disk.py --engine-url https://engine/ --username
>admin@internal --password-file password \
>    --cafile ca.pem --sd-name iscsi-1 --disk-format raw
>Fedora-Server-dvd-x86_64-30-1.2.iso
>
>$ python3 download_disk.py --engine-url https://engine/ --username
>admin@internal --password-file password \
>    --cafile ca.pem --format raw 5f0b5347-bbbc-4521-9ca0-8fc17670bab0
>iso.raw
>
>$ qemu-img compare iso.raw Fedora-Server-dvd-x86_64-30-1.2.iso
>Images are identical.
>
>This is not easy to use and requires knowledge about oVirt internals
>(128
>MiB alignment), so I guess we
>need to make this simpler.
>
>Nir

Hi Nir,

The RFE is  https://bugzilla.redhat.com/show_bug.cgi?id=1787906

I definitely did not know about the 128MB rounding.

Anyway, my idea is that we need to be able to verify that the  ISO locally is 
also the same on oVirt.
If I get latest RHEL 8.1  DVD and it has been somehow corrupted  -> I would 
like to know it immediately , not after I have been debugging issues on a fresh 
install for hours. I think you got my point.

Best Regards,
Strahil Nikolov
_______________________________________________
Users mailing list -- users@ovirt.org
To unsubscribe send an email to users-le...@ovirt.org
Privacy Statement: https://www.ovirt.org/site/privacy-policy/
oVirt Code of Conduct: 
https://www.ovirt.org/community/about/community-guidelines/
List Archives: 
https://lists.ovirt.org/archives/list/users@ovirt.org/message/QLK56AEQV6L52SICGXL3PFPZD4HT5U5C/

Reply via email to