Headers set by router are no longer accessible in request_headers_to_add #41794
Description
Title: Headers set by router are no longer accessible in request_headers_to_add
Description:
We were mapping x-envoy-expected-rq-timeout-ms header into a different one, which stopped working when we upgraded from v1.34.0 to v1.35.0
request_headers_to_add:
- header:
key: "x-timeout"
value: '%REQ(x-envoy-expected-rq-timeout-ms)%'
append_action: ADD_IF_ABSENTUsing dev images, I was able to narrow it down to these commits 0672c68...0c00c89. Those don't have any image associated with them so I wasn't able to pinpoint the exact commit which introduced this change, but I suspect it stopped working due to 4453ce1, since it changes the timing of when the finalizeRequestHeaders is called.
Reporting it as bug since I didn't find this behaviour change in the release notes nor in the PR/commit descriptions.
I was also able to verify that this affects other headers (e.g. x-envoy-attempt-count), so it's not specific to x-envoy-expected-rq-timeout-ms.
Repro steps:
Using the envoy.yaml and docker-compose.yaml I attached below, run the following commands, and observe the headers that are being logged.
docker compose up
curl http://localhost:10000/fooWhen using envoyproxy/envoy:v1.34.0 image (and envoyproxy/envoy:dev-0672c682568ea2745d7a1433ef5c47353e728ffc) I'm seeing x-envoy-expected-rq-timeout-ms correctly being mapped to x-timeout, whereas with envoyproxy/envoy:v1.35.0 (and envoyproxy/envoy:dev-0c00c8960e845c0b089039908f9692b6dd50ce1e) that no longer works.
Config:
# envoy.yaml
admin:
address:
socket_address:
address: 0.0.0.0
port_value: 9902
static_resources:
listeners:
- name: outbound_listener
address:
socket_address:
address: 0.0.0.0
port_value: 10000
filter_chains:
- filters:
- name: envoy.filters.network.http_connection_manager
typed_config:
"@type": type.googleapis.com/envoy.extensions.filters.network.http_connection_manager.v3.HttpConnectionManager
stat_prefix: outbound_http
route_config:
name: local_route
virtual_hosts:
- name: echo
domains: ["*"]
routes:
- match: { prefix: "/" }
route:
cluster: echo
request_headers_to_add:
# the behaviour that's broken with the upgrade to v1.35.0
- header:
key: "x-timeout"
value: '%REQ(x-envoy-expected-rq-timeout-ms)%'
append_action: ADD_IF_ABSENT
http_filters:
- name: envoy.filters.http.router
typed_config:
"@type": type.googleapis.com/envoy.extensions.filters.http.router.v3.Router
clusters:
- name: echo
type: STRICT_DNS
connect_timeout: 5s
load_assignment:
cluster_name: echo
endpoints:
- lb_endpoints:
- endpoint:
address:
socket_address:
address: "echo"
port_value: 8080# docker-compose.yaml
services:
echo:
image: mendhak/http-https-echo
envoy:
# image: envoyproxy/envoy:v1.34.0
image: envoyproxy/envoy:v1.35.0
volumes:
- ./envoy.yaml:/etc/envoy/envoy.yaml
ports:
- "10000:10000"
- "9902:9902"
depends_on:
- echo