Enable IPv6 Support for Provisioning

[hardys]

By setting PROVISIONING_IPV6=true we configure the provisioning
bridge with an ipv6 address only, to enable further testing
of single-stack ipv6 provisioning

Related PR's

• Use kustomize for deploying baremetal-operator baremetal-operator#324
• Support for local image building #104
• Changes needed for IPv6 ironic-image#107
• IPv6 support for inspector ironic-inspector-image#42
• Allow overriding the URL ironic-ipa-downloader#11
• Turn off globbing on curl ironic-ipa-downloader#10
• https://review.opendev.org/#/c/695546/
• https://review.opendev.org/#/c/692110/
• Sync ironic-common.sh with ironic-image ironic-inspector-image#43

[metal3-io-bot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: hardys

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [hardys]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[stbenjam]

Status:

• BMO can deploy with custom config map
• Sample config map with IPv6 config
• provisioning host httpd can now start and listens on IPv6 address with correct inspector URL configured (the cluster one not yet, we'll need to add the env var to the config map)

[stbenjam]

I've got the kustomize version of things working, and got the basics for local image support up.

Ironic API is also up and responding on IPv6 now.

[stbenjam]

The way I've been testing this is

export IRONIC_LOCAL_IMAGE=file:///home/notstack/ironic-image
export IRONIC_INSPECTOR_LOCAL_IMAGE=file:///home/notstack/ironic-inspector-image
export KUSTOMIZE_OVERLAY=$PWD/ipv6
export PROVISIONING_IPV6=true

With the PR's linked in the description checked out in those paths.

[derekhiggins]

Changes required in the IPA image to support IPv6
Bump up ipa-ip-lookup-attempts to 6 - https://review.opendev.org/#/c/695546/ (merged in master)
Add IPv6 support in dhcp-all-interfaces - https://review.opendev.org/#/c/692110/

[hardys]

Changes required in the IPA image to support IPv6
Bump up ipa-ip-lookup-attempts to 6 - https://review.opendev.org/#/c/695546/ (merged in master)
Add IPv6 support in dhcp-all-interfaces - https://review.opendev.org/#/c/692110/

These both landed, as did the ironic-image changes - @stbenjam would you like to rebase/squash the series now so we can start a final review/test pass?

I'm thinking we could land this iteration then work on switching the BMC ips to v6 as a follow-up?

[hardys]

/test-integration

[hardys]

/test-centos-integration

[hardys]

/test-integration

[hardys]

/test-centos-integration

[hardys]

@stbenjam I tested locally and I think the CI failures are real - I see the nodes fail to start with a libvirt error starting the domain - it can't find /usr/share/OVMF/OVMF_VARS.fd

Does that sound familiar?

[stbenjam]

That'd be the OVMF firmware when doing UEFI boot, are you testing ipv6?

[stbenjam]

Looks like we forgot to explicitly installed edk2-ovmf (EL) and ovmf (Ubuntu) packages, I've added them to the Ansible roles.

[stbenjam]

/test-integration

[stbenjam]

/test-centos-integration

[hardys]

Thanks, this worked for me locally with ipv4 so lets hope CI agrees :)

[stbenjam]

/test-integration

[stbenjam]

/test-centos-integration

[stbenjam]

Tests are green!!

[hardys]

Good to see the CI pass now, and this works fine locally on centos7 without PROVISIONING_IPV6 set, but when I enable ipv6 we fail to pxeboot the ramdisk.

I wonder if that's a centos7 vs el8 thing, I'll re-test on el8 to confirm tomorrow.

[hardys]

Ok I re-tested this on a freshly installed RHEL8 host and it worked, although the pxe boot of node-1 hung and only worked after a manual force-reset of the VM.

@stbenjam I think we should go ahead and merge this, then iterate on any remaining issues, but do you think we should either add a comment in the example config that this won't work on centos7, or raise an issue to track that?

I suspect we may want to add centos8 support and just say this is unsupported on centos7, otherwise we'll have to figure out what the problems are with the roms on centos7.

[stbenjam]

Ok I re-tested this on a freshly installed RHEL8 host and it worked, although the pxe boot of node-1 hung and only worked after a manual force-reset of the VM.

@stbenjam I think we should go ahead and merge this, then iterate on any remaining issues, but do you think we should either add a comment in the example config that this won't work on centos7, or raise an issue to track that?

I suspect we may want to add centos8 support and just say this is unsupported on centos7, otherwise we'll have to figure out what the problems are with the roms on centos7.

There weren't any CentOS 8 cloud images when I looked. Today it looks like there are vagrant images in http://cloud.centos.org/centos/8/, so maybe they're getting closer to getting them all built.

The reason CentOS 7 doesn't work is because the cloud images are BIOS only. FCOS and Ubuntu both have dual BIOS/UEFI images. Hopefully when CentOS 8 images come out they will be dual boot.

[hardys]

Ok I raised #149 to track getting some support for centos w/ipv6 - lets merge this and iterate on that as a follow-up.

[hardys]

/lgtm

[metal3-io-bot]

@hardys: you cannot LGTM your own PR.

In response to this:

/lgtm

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.

[russellb]

/lgtm