rpm: Update virtualization packages

[andreabolognani]

What this PR does / why we need it:

Updates virtualization packages. Specifically:

• libvirt 7.6.0-6 → 8.0.0-2
• QEMU 6.0.0-33 → 6.2.0-5

The libvirt.org/go/libvirt Go package is also updated to match.

Release note:

This version of KubeVirt includes upgraded virtualization technology based on libvirt 8.0.0 and QEMU 6.2.0.
Each new release of libvirt and QEMU contains numerous improvements and bug fixes.

[andreabolognani]

@alicefr the libguestfs dependencies have changed significantly in the latest package. I have tested the scenarios described in docs/guestfs.md without encountering any issues, but I'm not really confident that covers everything. Can you please take a look?

[andreabolognani]

At least some of the failures look legit.

One is caused by netcat being missing from the image - libvirt used to drag this in, but it no longer does. We could add it back manually, but maybe the test case could be reworked so that doing that is not necessary? Perhaps there are other uses of netcat outside of the testsuite.

I'm unable to reproduce one of the failures locally because

S [SKIPPING] [6.113 seconds]
[rfe_id:393][crit:high][vendor:cnv-qe@redhat.com][level:system][sig-compute] VM Live Migration
tests/migration_test.go:97
  Starting a VirtualMachineInstance 
  tests/migration_test.go:360
    live migration cancelation
    tests/migration_test.go:2208
      should be able to cancel a migration
      vendor/github.com/onsi/ginkgo/extensions/table/table.go:92
        [sig-storage][storage-req][test_id:2732] with RWX block disk and virtctl [It]
        vendor/github.com/onsi/ginkgo/extensions/table/table_entry.go:41

        Skip test when Block storage is not present

I'm unclear on what I need to do here, advice would be appreciated :)

I'll look into the whole thing more tomorrow.

[alicefr]

@andreabolognani I'll have a try but generally if you succeed with the examples, then it is what we expect to work with the setup.

At least some of the failures look legit.

One is caused by netcat being missing from the image - libvirt used to drag this in, but it no longer does. We could add it back manually, but maybe the test case could be reworked so that doing that is not necessary? Perhaps there are other uses of netcat outside of the testsuite.

I'm unable to reproduce one of the failures locally because

S [SKIPPING] [6.113 seconds]
[rfe_id:393][crit:high][vendor:cnv-qe@redhat.com][level:system][sig-compute] VM Live Migration
tests/migration_test.go:97
  Starting a VirtualMachineInstance 
  tests/migration_test.go:360
    live migration cancelation
    tests/migration_test.go:2208
      should be able to cancel a migration
      vendor/github.com/onsi/ginkgo/extensions/table/table.go:92
        [sig-storage][storage-req][test_id:2732] with RWX block disk and virtctl [It]
        vendor/github.com/onsi/ginkgo/extensions/table/table_entry.go:41

        Skip test when Block storage is not present

I'm unclear on what I need to do here, advice would be appreciated :)

I'll look into the whole thing more tomorrow.

You are missing the storage class that support block storage. In this case ceph, you need to set export KUBEVIRT_STORAGE="rook-ceph-default" before creating the cluster with make cluster-up then it is installed and you can check it with kubectl get sc

[alicefr]

Some of the missing packages are:

• e2fsprogs-*: this should be fine because we use the tool in the fixed appliance, so no need in the container
• dhcp-*: we don't need any network as we are not operating on VMs but only on disks
• dnf-*: this is fine because we don't need to install or fetch the dependencies
• findutlis: I'll readd this
• gdisk: this I'm not sure
• genisoimage: this either I'm not sure
• squashfs-tools: same as for e2fsprogs
• supermin : this is fine as we use a fixed appliance

I'll try to find out in which cases libguestfs uses gdisk and genisoimage

EDIT: they are listed in the package list for the appliance so I guess they are used there. The new image size has also reduced from 457MB to 392MB :)

[andreabolognani]

You are missing the storage class that support block storage. In this case ceph, you need to set export KUBEVIRT_STORAGE="rook-ceph-default" before creating the cluster with make cluster-up then it is installed and you can check it with kubectl get sc

@alicefr thanks, I've tried this and it worked as expected. Did you find this documented anywhere, or is it tribal knowledge? Either way, I couldn't reproduce the failure locally so I'm going to assume it was a fluke until proven otherwise.

findutlis: I'll readd this

I'll try to find out in which cases libguestfs uses gdisk and genisoimage

they are listed in the package list for the appliance so I guess they are used there.

So you think that findutils, gdisk and genisoimage should all be added back to the guestfs image? If they're used by libguestfs at runtime even when working with a pre-generated appliance, isn't it a packaging bug that they're not dragged in explicitly? We should discuss this with Rich.

For the time being, I believe I have fixed one of the test failures. More to come next week :)

[alicefr]

@andreabolognani I think we should only add findutils as standard tool but gdisk and genisoimage no. I don't think we have a good split for using a fixed appliance in centos 8.

About the env variable, yes I don't think that is documented :/. This script uses some of the variables in KubeVirt CI: https://github.com/kubevirt/kubevirt/blob/main/automation/test.sh

[andreabolognani]

I think we should only add findutils as standard tool but gdisk and genisoimage no.

Done.

I don't think we have a good split for using a fixed appliance in centos 8.

AFAICT CentOS Stream 8 and 9 both have the changes we need.

If they're used by libguestfs at runtime even when working with a pre-generated appliance, isn't it a packaging bug that they're not dragged in explicitly? We should discuss this with Rich.

https://bugzilla.redhat.com/show_bug.cgi?id=1989514#c19

[andreabolognani]

Shouldn't be an issue since a couple of tests are still failing, but just to be on the safe side

/hold

[andreabolognani]

I'm still investigating the issue with booting the Alpine Linux 3.7 image using the RHEL 8.6 machine type, but in the meantime I've brought all images into the current decade :)

/hold cancel

[andreabolognani]

Updated to fix a couple of tests that were relying on the exact amount of memory used by CirrOS, which has changed when moving to the latest version.

I'm still seeing failures for [sig-network] [rfe_id:694][crit:medium][vendor:cnv-qe@redhat.com][level:component]Networking Multiple virtual machines connectivity using bridge binding interface with a test outbound VMI should be able to reach [test_id:1542]the internet in prow but the test works fine locally, so I'm hoping it might have have been a temporary issue of the testing environment.

[andreabolognani]

Most of the failures are probably caused by the ongoing CI cluster upgrade, I assume.

@rmohr @vladikr can you please take a look? I think this should be ready to be merged now :)

[rmohr]

Most of the failures are probably caused by the ongoing CI cluster upgrade, I assume.

Probably. It looks weird.

/retest

[kubevirt-commenter-bot]

/retest-required
This bot automatically retries required jobs that failed/flaked on approved PRs.
Silence the bot with an /lgtm cancel or /hold comment for consistent failures.

[andreabolognani]

The ipv6 test is still not working since we don't have DHCP6 so the VM is not familiar with the DNS IP.
The test should be skipped if the cluster doesn't support IPv4.
Sorry for the noise.

@AlonaKaplan does the diff below look reasonable to you? I'm preparing a rebase and I can just squash it in.

diff --git a/tests/network/vmi_networking.go b/tests/network/vmi_networking.go
index ebac17309..1b0f30e25 100644
--- a/tests/network/vmi_networking.go
+++ b/tests/network/vmi_networking.go
@@ -300,12 +300,13 @@ var _ = SIGDescribe("[rfe_id:694][crit:medium][vendor:cnv-qe@redhat.com][level:c
 
        Context("VirtualMachineInstance with default settings", func() {
                It("[test_id:1542]should be able to reach the internet", func() {
+                       libnet.SkipWhenClusterNotSupportIpv4(virtClient)
                        outboundVMI := libvmi.NewCirros(
                                libvmi.WithInterface(libvmi.InterfaceDeviceWithMasqueradeBinding()),
                                libvmi.WithNetwork(v1.DefaultPodNetwork()),
                        )
                        outboundVMI = runVMI(outboundVMI)
-                       Expect(libnet.WithIPv6(console.LoginToCirros)(outboundVMI)).To(Succeed())
+                       tests.WaitUntilVMIReady(outboundVMI, console.LoginToCirros)
 
                        By("checking the VirtualMachineInstance can fetch via HTTP")
                        err := console.SafeExpectBatch(outboundVMI, []expect.Batcher{

[AlonaKaplan]

The ipv6 test is still not working since we don't have DHCP6 so the VM is not familiar with the DNS IP.
The test should be skipped if the cluster doesn't support IPv4.
Sorry for the noise.

@AlonaKaplan does the diff below look reasonable to you? I'm preparing a rebase and I can just squash it in.

diff --git a/tests/network/vmi_networking.go b/tests/network/vmi_networking.go
index ebac17309..1b0f30e25 100644
--- a/tests/network/vmi_networking.go
+++ b/tests/network/vmi_networking.go
@@ -300,12 +300,13 @@ var _ = SIGDescribe("[rfe_id:694][crit:medium][vendor:cnv-qe@redhat.com][level:c
 
        Context("VirtualMachineInstance with default settings", func() {
                It("[test_id:1542]should be able to reach the internet", func() {
+                       libnet.SkipWhenClusterNotSupportIpv4(virtClient)
                        outboundVMI := libvmi.NewCirros(
                                libvmi.WithInterface(libvmi.InterfaceDeviceWithMasqueradeBinding()),
                                libvmi.WithNetwork(v1.DefaultPodNetwork()),
                        )
                        outboundVMI = runVMI(outboundVMI)
-                       Expect(libnet.WithIPv6(console.LoginToCirros)(outboundVMI)).To(Succeed())
+                       tests.WaitUntilVMIReady(outboundVMI, console.LoginToCirros)
 
                        By("checking the VirtualMachineInstance can fetch via HTTP")
                        err := console.SafeExpectBatch(outboundVMI, []expect.Batcher{

Yes. And the VM should have bridge binding as the original test had.

outboundVMI := libvmi.NewCirros()

Sorry again for my previous misleading request.

[andreabolognani]

Sorry again for my previous misleading request.

No worries, thank you for looking into it and helping out :)

[AlonaKaplan]

/lgtm

[andreabolognani]

/retest

[kubevirt-bot]

@andreabolognani: The following test failed, say /retest to rerun all failed tests or /retest-required to rerun all mandatory failed tests:

Test name	Commit	Details	Required	Rerun command
pull-kubevirt-e2e-kind-1.19-vgpu	3883bff	link	true	/test pull-kubevirt-e2e-kind-1.19-vgpu

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. I understand the commands that are listed here.

[kubevirt-commenter-bot]

/retest-required
This bot automatically retries required jobs that failed/flaked on approved PRs.
Silence the bot with an /lgtm cancel or /hold comment for consistent failures.

[EdDev]

/test pull-kubevirt-e2e-kind-1.22-sriov