bryancall opened a new pull request, #13915: URL: https://github.com/apache/trafficserver/pull/13915
Both proxies in `test_proxy_serve_stale_dns_fail` point `proxy.config.dns.nameservers` at the test's own DNS server, but the test leaves `proxy.config.dns.resolv_conf` at its default. ATS then also adds the nameservers from the host's `/etc/resolv.conf`, so some lookups for the test's unresolvable origin go to whatever resolver the CI host has. Of the 47 Uranium tests that set `dns.nameservers`, this is the only one that does not also set `resolv_conf` to `NULL`. When the host's resolver is slow, the lookup stalls. Before #13869 that made the test serve a `500 Cannot find server.` where it expected a stale `200`, which is how it failed in builds 45685, 45763 and 45891. #13869 widened the stale window, so a single stall no longer fails it, but a host resolver that does not answer still runs the test past its 70 second limit. This sets `resolv_conf` to `NULL` on both proxies, so the test only talks to its own DNS server. ## Testing On hera (Fedora 44) at master 7b9a5559dc, I pointed `resolv_conf` at a local resolver standing in for the host's, then ran the test with that resolver behaving three ways: | Test | Host resolver | Result | |---|---|---| | master, unchanged | hera's own | 3/3 pass | | before #13869 | stalls the first lookup 30 s | 3/3 fail, `500` with `Age: 40` as in 45685 | | master | stalls the first lookup 30 s | 5/5 pass | | master | never answers | 3/3 fail, 70 s timeout | | this change | never answers (still listening) | 10/10 pass, no queries reached it | -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
