Proposal to Phase Out Cirros in Favor of Alpine for Test Images

[Vamsikrishna Siddu]

Hello Everyone,

Iam starting this discussion based on feedback I received in a recent PR of mine by akal...@redhat.com. https://github.com/kubevirt/kubevirt/pull/15328#issuecomment-3333355969. Where I am replacing cirros with alpine for couple of the E2E storage tests.That review comment raised a broader question: even if we update existing tests, we may still end up with new PRs adding Cirros-based tests unless we define a clear guideline.To avoid this recurring cycle, I’d like to propose that we stop using Cirros for any new tests going forward. 

issue link: https://github.com/kubevirt/kubevirt/issues/15043

Regards,

Siddu Vamsikrishna,

Software Engineer,

IBM R&D Bangalore,

[Felix Matouschek]

Hi,

Am I correct in understanding that this is primarily motivated by Cirros being incompatible with s390x?

I wanted to check if we have settled on Alpine as the new baseline, or if we are still exploring other minimal image solutions?

My main concern is that moving away from Cirros might increase the resource footprint during our e2e runs. Since Cirros is quite lightweight, have we evaluated the potential impact of Alpine (or other alternatives) on our CI infrastructure load?

Thanks,
Felix

[Edward Haas]

On Tue, Dec 2, 2025 at 12:35 PM 'Felix Matouschek' via kubevirt-dev <kubevi...@googlegroups.com> wrote:

Hi,

Am I correct in understanding that this is primarily motivated by Cirros being incompatible with s390x?

I wanted to check if we have settled on Alpine as the new baseline, or if we are still exploring other minimal image solutions?

My main concern is that moving away from Cirros might increase the resource footprint during our e2e runs. Since Cirros is quite lightweight, have we evaluated the potential impact of Alpine (or other alternatives) on our CI infrastructure load?

The memory footprint assigned for both are the same. What resources do you think may be affected? 

Thanks,
Felix

On Tuesday, December 2, 2025 at 10:26:52 AM UTC+1 Vamsikrishna Siddu wrote:

Hello Everyone,

Iam starting this discussion based on feedback I received in a recent PR of mine by akal...@redhat.com. https://github.com/kubevirt/kubevirt/pull/15328#issuecomment-3333355969. Where I am replacing cirros with alpine for couple of the E2E storage tests.That review comment raised a broader question: even if we update existing tests, we may still end up with new PRs adding Cirros-based tests unless we define a clear guideline.To avoid this recurring cycle, I’d like to propose that we stop using Cirros for any new tests going forward. 

Even if we agree to prefer Alpine, a guideline is not enough to enforce it in practice.

You will need reviewers and approvers to remember this, which IMO is hard to assure.

Have you considered replacing all usages of Cirros and then dropping the image and factory from the source?

Even if new tests are introduced with Cirros, you will reach them eventually.

 

issue link: https://github.com/kubevirt/kubevirt/issues/15043

Regards,

Siddu Vamsikrishna,

Software Engineer,

IBM R&D Bangalore,

--
You received this message because you are subscribed to the Google Groups "kubevirt-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kubevirt-dev...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/kubevirt-dev/39102241-31a2-4e09-9e31-9c807c257781n%40googlegroups.com.

[Vamsikrishna Siddu]

Hello Edward,

Have you considered replacing all usages of Cirros and then dropping the image and factory from the source?

Even if new tests are introduced with Cirros, you will reach them eventually.

For the above question,  We cannot replace all the cirros tests with alpine. Because alpine does not have the tooling  required for the storage and network tests. For those tests I am thinking of alpinetest tooling as an alternative for cirros.

[Felix Matouschek]

Am Dienstag, dem 02.12.2025 um 13:06 +0200 schrieb Edward Haas:

On Tue, Dec 2, 2025 at 12:35 PM 'Felix Matouschek' via kubevirt-dev <kubevi...@googlegroups.com> wrote:

Hi,

Am I correct in understanding that this is primarily motivated by Cirros being incompatible with s390x?

I wanted to check if we have settled on Alpine as the new baseline, or if we are still exploring other minimal image solutions?

My main concern is that moving away from Cirros might increase the resource footprint during our e2e runs. Since Cirros is quite lightweight, have we evaluated the potential impact of Alpine (or other alternatives) on our CI infrastructure load?

The memory footprint assigned for both are the same. What resources do you think may be affected? 

Right, I missed that.

Then it's only a matter of how long it takes for Alpine to boot in comparison to Cirros to not increase the run times of lanes.

[Vamsikrishna Siddu]

Hi Felix,

Right, I missed that.

Then it's only a matter of how long it takes for Alpine to boot in comparison to Cirros to not increase the run times of lanes.

Actually, I have compared the run times of lanes after replacing cirros to alpine.  What i have observed is that even after replacing the cirros to alpine. There is no increase in the test time run. I have put the details in this issue here. https://github.com/kubevirt/kubevirt/issues/15043#issuecomment-3096482834

I also analyzed the count of tests which i cannot replace cirros with alpine for the different sigs. https://github.com/kubevirt/kubevirt/issues/15043#issuecomment-3100670300

I am thinking of alpineTestTooling as an alternative for those tests.

[Felix Matouschek]

On Wednesday, December 3, 2025 at 6:06:23 AM UTC+1 Vamsikrishna Siddu wrote:

Hi Felix,

Right, I missed that.

Then it's only a matter of how long it takes for Alpine to boot in comparison to Cirros to not increase the run times of lanes.

Actually, I have compared the run times of lanes after replacing cirros to alpine.  What i have observed is that even after replacing the cirros to alpine. There is no increase in the test time run. I have put the details in this issue here. https://github.com/kubevirt/kubevirt/issues/15043#issuecomment-3096482834

I also analyzed the count of tests which i cannot replace cirros with alpine for the different sigs. https://github.com/kubevirt/kubevirt/issues/15043#issuecomment-3100670300

I am thinking of alpineTestTooling as an alternative for those tests.

Sounds good to me then! Thanks for testing it.