Enable unicast keepalived for all on-prem platforms #3016
Conversation
Initially this was only enabled for baremetal in the interest of limiting the impacts of any new bugs introduced by the change. We've now been running with unicast on baremetal since 4.6 and should have worked out the kinks. Additionally, other on-prem platforms (notably vsphere) are running into similar problems to what prompted the change on baremetal. In the interest of simplifying the support matrix for on-prem platforms, this enables unicast on all of them. For the moment that entails only hard-coding the ENABLE_UNICAST env var and moving the flip-mode file template from baremetal-specific to all of on-prem. I'm leaving the render functions for the moment in the interest of avoiding issues on upgrade if an old template gets rendered by a newer MCO. They can be cleaned up in 4.12.
seems to be the reason all the jobs insta-failed. I am confused why make update made no changes here. Was there a change to how the manifests directory is handled? |
|
/retest |
|
/approve |
openshift-ci
bot
added
the
approved
|
@cybertron what is the plan for upgrade? Is that an issue? |
There is code for handling the transition to unicast on upgrade, but we discovered that it has a bug[0] as part of the discussion that led to this change. However, all that means is existing clusters stay multicast and new clusters get unicast. Since clusters that were installed multicast were obviously working in that mode, I don't see it as a blocker to switching new clusters (although we do want to fix it for consistency, of course). We wanted to push ahead with this change first so we can get a lot of soak time on the change in the 4.11 cycle just in case there are any platform-specific issues that come up. If we can fix the mode change bug in 4.11 too then that's a bonus, but either way this will fix the vrid collision problem for new clusters. |
|
/test e2e-metal-ipi |
|
/test e2e-vsphere |
mandre
left a comment
mandre
left a comment
There was a problem hiding this comment.
LGTM as far as OpenStack is concerned.
|
vSphere install and upgrade failure unrelated /lgtm |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: cgwalters, cybertron, jcpowermac The full list of commands accepted by this bot can be found here. The pull request process is described here DetailsNeeds approval from an approver in each of these files:
Approvers can indicate their approval by writing |
|
/retest-required Please review the full test history for this PR and help us cut down flakes. |
1 similar comment
|
/retest-required Please review the full test history for this PR and help us cut down flakes. |
|
/test e2e-vsphere-upgrade |
|
/retest-required Please review the full test history for this PR and help us cut down flakes. |
|
@cybertron: The following tests failed, say
Full PR test history. Your PR dashboard. DetailsInstructions 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. |
This isn't used anymore since we enabled Unicast for all on-prem platforms which use Keepalived, via openshift#3016.
7 participants
Initially this was only enabled for baremetal in the interest of
limiting the impacts of any new bugs introduced by the change. We've
now been running with unicast on baremetal since 4.6 and should have
worked out the kinks. Additionally, other on-prem platforms (notably
vsphere) are running into similar problems to what prompted the
change on baremetal.
In the interest of simplifying the support matrix for on-prem
platforms, this enables unicast on all of them. For the moment that
entails only hard-coding the ENABLE_UNICAST env var and moving the
flip-mode file template from baremetal-specific to all of on-prem.
I'm leaving the render functions for the moment in the interest of
avoiding issues on upgrade if an old template gets rendered by a
newer MCO. They can be cleaned up in 4.12.
- What I did
- How to verify it
- Description for the changelog
Enable unicast keepalived for all on-prem platforms.