Some e2e tests requiring SCTP to be enabled failed on the origin rebase PR, prompting tests with the [Feature:SCTPConnectivity] tag to be marked for skip in origin's rules.go. Since this feature is beta and enabled by default in 1.19, this tests should either be fixed or the [Disabled:Alpha] skip switched to a [Disabled:SpecialConfig] skip.
Bugs aside, to have the tests pass the SCTP kernel module loaded on the nodes, which I don't think is happening on the current ci. See https://docs.openshift.com/container-platform/4.5/networking/using-sctp.html#nw-sctp-enabling_using-sctp
origin #25818 adds support for being able to un-skip the tests, but does not unskip them in any existing e2e jobs. We still need to figure out what to do there. I may add a periodic.
*** Bug 1908677 has been marked as a duplicate of this bug. ***
(In reply to Dan Winship from comment #2) > origin #25818 adds support for being able to un-skip the tests, but does not > unskip them in any existing e2e jobs. We still need to figure out what to do > there. I may add a periodic. So that merged, so now it's _possible_ to run the SCTPConnectivity tests; someone needs to create an e2e job where 1. There's a MachineConfig to load sctp.ko, as in https://access.redhat.com/solutions/4727321 2. openshift-tests gets run with --provider '{"type": "aws", "hasSCTP": true }' (I don't think GCE supports SCTP over the underlay network at all, so I think we have to run the SCTP tests on AWS.) With that provider override, openshift-tests will include the SCTPConnectivity tests in the set of tests that get run. Actually, I guess we probably want 2 such e2e jobs; one for openshift-sdn and one for ovn-kubernetes. Probably as a nightly period job, and then also with not-run-by-default presubmit copies in sdn, ovn-kubernetes, and cluster-network-operator.
@rravaiol, here[0] is a small slack thread discussing this more. I'm not sure when, if ever, I would get to this unless we want to bump it's priority for some reason. Sounds like QE might get coverage of this already. We could find a way to get it running in some existing e2e CI jobs if needed. For now this is medium priority and I'll try to get to it at some point, if possible.
(In reply to jamo luhrsen from comment #10) > @rravaiol, here[0] is a small slack thread discussing this more. > I'm not sure when, if ever, I would > get to this unless we want to bump it's priority for some reason. Sounds > like QE might get coverage of this > already. We could find a way to get it running in some existing e2e CI jobs > if needed. For now this is medium > priority and I'll try to get to it at some point, if possible. [0] https://coreos.slack.com/archives/CDCP2LA9L/p1667408550662919