Ci improvements #1500
Ci improvements #1500
Conversation
|
/assign @dave-tucker |
|
/retest |
|
/retest |
|
hmm only the HA tests are failing with etcd timeouts hopefully we can get the new 3.4.10 version in kind soon with the unsafe fsync feature that will make it more suitable for testing etcd-io/etcd#11946 (cc @girishmg @dcbw ) |
|
This runs
/assign @dave-tucker @dcbw @girishmg |
build the ovn-kube image only once Signed-off-by: Antonio Ojea <[email protected]>
it panics, we have to wait until kubernetes/kubernetes#93119 and backport to the corresponding branch Signed-off-by: Antonio Ojea <[email protected]>
Kubernetes upstream has a helper script to run ginkgo, that allows to can run tests in parallel, that will improve the time to run the jobs by several orders of magnitude. From 50 mins to 7 mins to run the whole sig-network tests This way we don't neet to shard the tests in several jobs. It also allows to configure retrying flake tests, We are already retrying PRs, this at least will give signal on totally broken tests, if a test is not working after X times, that's something we should investigate. Signed-off-by: Antonio Ojea <[email protected]>
| # all tests that have P as the sixth letter after the N | ||
| GINKGO_ARGS="${GINKGO_ARGS} "'--ginkgo.focus=\[sig-network\]\s[Nn].{6}[Pp].*$' | ||
| shard-conformance) | ||
| FOCUS="\\[Conformance\\]" |
It allows to run tests in parallel, reducing the time of the job and adding more stress to OVN so we can test its performance. Currently set to 4 tests in parallel Signed-off-by: Antonio Ojea <[email protected]>
…2-fix_ds_acl_cluster OCPBUGS-6823: [release-4.12] Fix Egress FW ACL rules in dualstack mode
- What this PR does and why is it needed
It runs all the [sig-network] tests in the same job, removing current sharding. We are constrained on number of parallel jobs and the new features we are adding, require new independent jobs.
It allows to run the tests in parallel, right now we run with 4 tests in parallel, running a job in 20 minutes less
It build the ovn-kubernetes image only once, and share it using github actions artifacts with the multiple jobs.
- Special notes for reviewers
- How to verify it
- Description for the changelog