Kong upstream health flip-flopping due to TTL=0 handling

[onematchfox]

Summary

Kong version: 2.4.1

See Kong/lua-resty-dns-client#131

Apologies for the duplicate issue. Opened the issue in the library in which I believe the bug lies. However, from experience it would seem like the maintainers may not be watching the underlying libraries as closely as this one. So, given that the issue hasn't gotten any attention despite being open for over a month and, since the issue ultimately manifests itself within Kong, I thought perhaps it would be better if I raised it here.

[chris-branch]

Is there any update on this issue? I am seeing the same issue with an upstream that is hosted in Azure using Azure's DNS, which appears to return TTL = 0 periodically. I put together a small, standalone example that reproduces the issue fairly consistently. See the README in this repo:

https://github.com/chris-branch/kong-dnsbug-2

Also, #7551 is likely the root cause of #8056
Edit: 8056 may be a separate issue based on new info added over there.

[chris-branch]

@Tieske - Any thoughts on whether the proposed solution outlined in Kong/lua-resty-dns-client#131 is reasonable and could be included in a future release?

[hbagdi]

Please try out Kong 3.2 or above and report back if you still run into this problem.