Add udp-lb testcase for amphora provider #56
Conversation
rlobillo
force-pushed
the
udplb
branch
2 times, most recently
from
c3a4a0f
to
b00a8bf
Compare
|
/retest |
|
/retest |
2 similar comments
|
/retest |
|
/retest |
rlobillo
force-pushed
the
udplb
branch
3 times, most recently
from
831870e
to
f31ffd8
Compare
|
/retest |
|
/retest |
| // Check if the provided lbMethod is applied considering the | ||
| // result map (<pod name>: <number of accesses>) and the list of pods provided. | ||
| // TO DO: Only 'ROUND_ROBIN' method implement for the moment. | ||
| func isLbMethodApplied(lbMethod string, results map[string]int, pods *v1.PodList) bool { |
There was a problem hiding this comment.
I don't exactly see this testing other algorithms. And it feels like something that may be pretty flaky, even if we have this safeguard of 20. It's probably necessary to consult Octavia engineers on how reliable the algorithm choice is. I would consider just checking if pool was created with a correct setting instead.
There was a problem hiding this comment.
I want to validate the end to end functionality here... I think that checking the pool can be tricky if something change on openstack. Our plan is to run this on D/S in all the supported OSP releases so I think it is more stable this high level check than checking the pool.
About other algorithms, I think the most probable algorithm to include here is SOURCE_IP_PORT on OVN, but we agreed that monitors are needed and it is currently not supported in OVN provider. I will mention that in separate PR where I will include the test for OVN provider.
| o.Expect(pool.LBMethod).Should(o.Equal(lbMethod), "Unexpected LBMethod on Openstack Pool: %q", pool.LBMethod) | ||
|
|
||
| g.By("accessing the service 100 times from outside and storing the name of the pods answering") | ||
| time.Sleep(10 * time.Minute) // Give time to the cluster to be ready |
There was a problem hiding this comment.
Wow, that's pretty long time. What are we actually waiting for here?
There was a problem hiding this comment.
I am still struggling with vexxhost to get the results from the pods. It is timing out sometimes. I added this long sleep (10 minutes) and the last gate run passed, so I am updating the check to allow timeouts in the queries, and reducing this timer just to see if there is no need to wait so long.
There was a problem hiding this comment.
Alright, let's see. Obviously a wait for some condition instead of Sleep is better, so we might want to consider what we're waiting for.
There was a problem hiding this comment.
Finally there is no need to wait anything. Just accepting some timeouts in the 100 requests is enough to provide stability on the test.
The testing for this patch is done!
|
/retest |
1 similar comment
|
/retest |
|
@rlobillo: all tests passed! Full PR test history. Your PR dashboard. 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. I understand the commands that are listed here. |
|
@dulek please help to get this merged :) I'd like to include more tests based on this one. |
|
/lgtm Looks good now, but I'd rather ask either @MaysaMacedo or @pierreprinetti to give final |
| o.Expect(err).NotTo(o.HaveOccurred()) | ||
| o.Expect(pool.Protocol).Should(o.Equal("UDP"), "Unexpected protocol on Openstack LoadBalancer Pool: %q", pool.Name) | ||
| //Set as OFFLINE in vexxhost despite the lb is operative | ||
| //o.Expect(pool.OperatingStatus).Should(o.Equal("ONLINE"), "Unexpected Operating Status on Openstack Pool: %q", pool.Name) |
There was a problem hiding this comment.
was a ticket open about this to vexxhost?
| results[podName]++ | ||
| } | ||
| } | ||
| e2e.Logf("Pods accessed after 100 UDP requests:\n%v\n", results) |
There was a problem hiding this comment.
where could I find this log in the results?
I tried checking here https://gcsweb-ci.apps.ci.l2s4.p1.openshiftapps.com/gcs/origin-ci-test/pr-logs/pull/openshift_openstack-test/56/pull-ci-openshift-openstack-test-main-test/1584437086490988544/artifacts/test/openstack-test-openstack/artifacts/e2e.log
There was a problem hiding this comment.
That logs are only appearing if you run the test isolated:
KUBECONFIG=~/.kube/config ./openstack-tests run-test "[sig-installer][Suite:openshift/openstack][lb] The Openstack platform should create an UDP Amphora LoadBalancer when an UDP svc with type:LoadBalancer is created on Openshift [Suite:openshift/conformance/parallel]"
or if you run the whole testsuite but the test fails.
|
Great work! |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: MaysaMacedo 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:
Approvers can indicate their approval by writing |
openshift-ci
bot
added
the
approved
Covers partially the epic https://issues.redhat.com/browse/OSASINFRA-2753
This test creates an UDP lb on Openshift and checks the created Openstack resources. It also sends 100 requests to the LoadBalancer Service and analyze the pods' responses so that the e2e functionality is verified.
This test will be skipped for Kuryr NetworkType and if cloudProviderConfig is configuring a different lb-provider than Amphora.