no healthy upstream when simultaneously creating Service+Ingress

[clementnuss]

Bug Description

I am currently testing/comparing a series of Ingress Controllers, and for that matter am testing the behaviour of the controllers in dynamic Ingress/Service environments.

A test round unfolds as follows: a bash script automatically creates 50 Services+Ingresses, runs a series of k6 load tests while there is no change made to the Ingress definitions and then another series of k6 tests while randomly modifying the ingresses (shuffling backends, deleting/re-creating). Finally, the script deletes the namespace containing the test deployment (and therefore the services and the ingresses)

If I run the same test round / bash script directly afterwards, I get a lot of no healthy upstream response from Istio. Instead of 100% of 200/OK HTTP answers, I get ~40% of 503/No healthy upstream responses. The only solution then is to kill istiod and ìstio-ingressgateway`, then everything is perfectly fine again !

I think this issue is related to #35172 and #35293 , and I have observed that when I create my Services, sleep 1, and then create the Ingresses, I do not get the issue I just described and Istio is working perfectly fine, even after several creation/tests/deletions rounds

Version

❯ istio-1.11.2/bin/istioctl --istioNamespace kube-istio version
client version: 1.11.2
control plane version: 1.11.2
data plane version: 1.11.2 (5 proxies)
❯ kubectl version --short
Client Version: v1.22.2
Server Version: v1.21.4

Additional Information

No response

[ramaraochavali]

This seems to be the same issue as #35293. Sorry for the regression. This #35298 fixes it for 1.11

[istio-policy-bot]

🚧 This issue or pull request has been closed due to not having had activity from an Istio team member since 2021-09-22. If you feel this issue or pull request deserves attention, please reopen the issue. Please see this wiki page for more information. Thank you for your contributions.

Created by the issue and PR lifecycle manager.