You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
gateway: the dns01 migration path in the README restores the ambiguity, not termination #4197
Split out of the review of #3342, carried as a follow-up rather than blocking the release.
packages/extra/gateway/README.md offers three ways forward for anyone depending on the terminate path for one of the shed hostnames. The second one reads as more than it delivers: "Switching certMode to dns01 or existingSecret keeps the pre-upgrade shape wholesale".
That is literally true, and the trouble is what the pre-upgrade shape actually was. rendersTLSPassthrough (internal/controller/tenantgateway/reconciler.go:94-96) excludes edge alone, and both passthrough render loops sit outside the certMode switch that ends at reconciler.go:1585. So under dns01 the tls-<svc> listeners still stand beside a wildcard terminate listener covering the same names — which is exactly the two-chains-on-one-name ambiguity the README explains elsewhere that you cannot observe from the cluster, since the pinned Cilium has no Conflicted condition.
Someone picking that path to get Gateway termination back therefore gets the undetermined race back instead, not termination. Worth saying so where the path is offered, so the three options can be compared honestly: the first (move the consumer to another hostname under the apex) is the one that actually gives up nothing.
Fixed in #3342 at 8af98639. packages/extra/gateway/README.md:109 now says the pre-upgrade shape is the caveat rather than the benefit: those modes leave the tls-<svc> listeners standing beside the wildcard terminate listener covering their names, both port-443 chains carry the same SNI again, and nothing on the objects says which one answers.
Activity
kvaps commentedon Sep 10, 2026
Fixed in #3342 at
8af98639.packages/extra/gateway/README.md:109now says the pre-upgrade shape is the caveat rather than the benefit: those modes leave thetls-<svc>listeners standing beside the wildcard terminate listener covering their names, both port-443 chains carry the same SNI again, and nothing on the objects says which one answers.