Bug 1740193 - CDI: import from http fails - never finishes
Summary: CDI: import from http fails - never finishes
Keywords:
Status: CLOSED ERRATA
Alias: None
Product: Container Native Virtualization (CNV)
Classification: Red Hat
Component: Storage
Version: 2.1.0
Hardware: Unspecified
OS: Unspecified
unspecified
urgent
Target Milestone: ---
: 2.1.0
Assignee: Alexander Wels
QA Contact: Natalie Gavrielov
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2019-08-12 12:50 UTC by Natalie Gavrielov
Modified: 2019-10-31 14:09 UTC (History)
8 users (show)

Fixed In Version:
Doc Type: If docs needed, set a value
Doc Text:
Clone Of:
: 1744602 (view as bug list)
Environment:
Last Closed: 2019-10-31 14:08:47 UTC
Target Upstream Version:
Embargoed:


Attachments (Terms of Use)


Links
System ID Private Priority Status Summary Last Updated
Github kubevirt containerized-data-importer pull 915 0 'None' closed Fallback to Fedora29 minimal for images 2020-12-21 07:04:47 UTC
Red Hat Product Errata RHEA-2019:3282 0 None None None 2019-10-31 14:09:01 UTC

Comment 2 Natalie Gavrielov 2019-08-14 12:32:53 UTC
Hi Alexander,

This issue impacts upstream only since qemu-img is taken from fedora, and no impact on downstream (since qemu there is taken from rhel), right?

Comment 3 Alexander Wels 2019-08-14 12:37:28 UTC
The only one we know for certain that is broken is qemu 3.1.X upstream. If there are downstream builds based on that, then those would be broken as well. But for now we should assume downstream works, unless proven otherwise.

Comment 4 Gowrishankar Rajaiyan 2019-08-14 12:52:44 UTC
We should re-assign this bug to the qemu component.

Comment 5 Alexander Wels 2019-08-14 12:56:18 UTC
This is already fixed for qemu 4.0.x its just that fedora 30 doesn't include the 4 series yet.

Comment 6 Gowrishankar Rajaiyan 2019-08-18 04:32:09 UTC
Ok, then this should be marked as a duplicate of the qemu bug.

Comment 7 Ying Cui 2019-08-21 12:25:25 UTC
Could we catch this issue on CDI upstream to make sure we have test case to cover it.

Comment 8 Alexander Wels 2019-08-21 12:27:10 UTC
We have added a functional test using a known image that triggered the issue. So if it happens again that test should start failing.

Comment 10 Ademar Reis 2019-08-22 12:27:03 UTC
(In reply to Alexander Wels from comment #3)
> The only one we know for certain that is broken is qemu 3.1.X upstream. If
> there are downstream builds based on that, then those would be broken as
> well. But for now we should assume downstream works, unless proven otherwise.

Can you please provide a pointer to a public image that triggers this issue? We would like to review the problem to understand which versions are affected and have a test in RHEL as well.

Comment 11 Alexander Wels 2019-08-22 12:36:47 UTC
Sure, so the combination of things that need to happen is pretty specific. You have to do a streaming conversion from an http source. If you download the file to a local location and then convert it works.

qemu-img convert -p -O raw https://download.cirros-cloud.net/0.4.0/cirros-0.4.0-x86_64-disk.img

That particular one will start converting and usually hang around 4.7% progress. I have heard other people mention any qcow2 from an http source will exhibit the same behaviour. So this appears to happen with the 3.1.z series of qemu. We had to downgrade the version we were using to 3.0.z series in our upstream releases. We have also verified that the 4.0.0 version works as well. But we had no viable source to get it from easily (maybe Fedora 31?)

In short for 3.1.z version qemu any qcow2 from http source will display the problem, but the cirros image in small and easily usable in tests.

Comment 12 Ademar Reis 2019-08-22 12:57:51 UTC
(In reply to Alexander Wels from comment #11)
> Sure, so the combination of things that need to happen is pretty specific.
> You have to do a streaming conversion from an http source. If you download
> the file to a local location and then convert it works.
> 
> qemu-img convert -p -O raw
> https://download.cirros-cloud.net/0.4.0/cirros-0.4.0-x86_64-disk.img
> 
> That particular one will start converting and usually hang around 4.7%
> progress. I have heard other people mention any qcow2 from an http source
> will exhibit the same behaviour. So this appears to happen with the 3.1.z
> series of qemu. We had to downgrade the version we were using to 3.0.z
> series in our upstream releases. We have also verified that the 4.0.0
> version works as well. But we had no viable source to get it from easily
> (maybe Fedora 31?)
> 

We have a "virt-preview" repository for Fedora which gets frequent upstream rebases. It's less stable, but might be useful: https://fedoraproject.org/wiki/Virtualization_Preview_Repository

> In short for 3.1.z version qemu any qcow2 from http source will display the
> problem, but the cirros image in small and easily usable in tests.

This is useful info, thank you. I'll clone this BZ and ask our QE to try to reproduce it and work on a test to avoid regressions.

Comment 13 Natalie Gavrielov 2019-09-25 07:37:06 UTC
Verified in CNV 2.1 - the described issue wasn't reproduced in downstream.

Comment 15 errata-xmlrpc 2019-10-31 14:08:47 UTC
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHEA-2019:3282


Note You need to log in before you can comment on or make changes to this bug.