Skip to content

fix: MaxStreamDuration should worked for non-route cluster#9028

Merged
zhaohuabing merged 3 commits into
envoyproxy:mainfrom
zirain:fix/maxStreamDuration
May 29, 2026
Merged

fix: MaxStreamDuration should worked for non-route cluster#9028
zhaohuabing merged 3 commits into
envoyproxy:mainfrom
zirain:fix/maxStreamDuration

Conversation

@zirain

@zirain zirain commented May 19, 2026

Copy link
Copy Markdown
Member

This seems to be missed in #7194

cc @arkodg

@zirain zirain requested a review from a team as a code owner May 19, 2026 00:57
@netlify

netlify Bot commented May 19, 2026

Copy link
Copy Markdown

Deploy Preview for cerulean-figolla-1f9435 ready!

Name Link
🔨 Latest commit 901599b
🔍 Latest deploy log https://app.netlify.com/projects/cerulean-figolla-1f9435/deploys/6a191a4c68306c0008a2cf6a
😎 Deploy Preview https://deploy-preview-9028--cerulean-figolla-1f9435.netlify.app
📱 Preview on mobile
Toggle QR Code...

QR Code

Use your smartphone camera to open QR code link.

To edit notification comments on pull requests, go to your Netlify project configuration.

@codecov

codecov Bot commented May 19, 2026

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 74.78%. Comparing base (560a262) to head (901599b).

Additional details and impacted files
@@            Coverage Diff             @@
##             main    #9028      +/-   ##
==========================================
- Coverage   74.78%   74.78%   -0.01%     
==========================================
  Files         252      252              
  Lines       40600    40612      +12     
==========================================
+ Hits        30364    30370       +6     
- Misses       8163     8168       +5     
- Partials     2073     2074       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
  • 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

@zhaohuabing

zhaohuabing commented May 28, 2026

Copy link
Copy Markdown
Member

Question on the semantics here: this change makes the same BackendTrafficPolicy timeout.http.maxStreamDuration value land in both the route action and the upstream cluster common HTTP protocol options.

Should route-scoped backend traffic rely on the route-level RouteAction.max_stream_duration instead, and leave the cluster-level commonHttpProtocolOptions.max_stream_duration unset for those generated route clusters? The route field seems to map more directly to a BackendTrafficPolicy targeting an HTTPRoute/rule, while the cluster field is backend/upstream scoped and introduces a second timer with the same value but slightly different scope.

I can see cluster-level max_stream_duration being useful for internal/non-route upstream clusters such as ALS, OTel, OIDC/JWKS, ext-auth, etc., where there is no route action to carry the timeout. Is the intent here to enforce the same value at both route and upstream-cluster level for normal routed traffic, or should this be limited to cases that do not already have a route-level max stream duration?

@zirain

zirain commented May 28, 2026

Copy link
Copy Markdown
Member Author

this's more like downstream vs upstream depends on where you specify it.
IIUC, for backendSettings it should result as part of cluster.

@zirain zirain force-pushed the fix/maxStreamDuration branch from 8aaa59d to 70fa97c Compare May 28, 2026 02:00

@zhaohuabing zhaohuabing left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

IMO this should only apply to non-route/internal clusters, where there is no RouteAction available to carry the timeout. For clusters generated from HTTPRoute/GRPCRoute backend traffic, we should rely on the route-level RouteAction.max_stream_duration instead. Otherwise we end up configuring two timers with the same value but different scopes: one scoped to the matched route, and one scoped to the upstream cluster.

@zirain

zirain commented May 28, 2026

Copy link
Copy Markdown
Member Author

IMO this should only apply to non-route/internal clusters, where there is no RouteAction available to carry the timeout. For clusters generated from HTTPRoute/GRPCRoute backend traffic, we should rely on the route-level RouteAction.max_stream_duration instead. Otherwise we end up configuring two timers with the same value but different scopes: one scoped to the matched route, and one scoped to the upstream cluster.

When using a BTP target to HTTPRoute, the value is same on RouteAction.max_stream_duration and the cluster.

@zhaohuabing

Copy link
Copy Markdown
Member

On second thought, the cluster-level max_stream_duration is effectively a no-op when both the route-level and cluster-level are set as the same value, so this is non-blocking.

It would still be good to add a comment around the cluster-level max_stream_duration to explain what EG is doing here, and clarify that the cluster-level setting has no effect since the same value is configured at both levels.

zhaohuabing
zhaohuabing previously approved these changes May 28, 2026

@zhaohuabing zhaohuabing left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Non-blocking: I’d prefer not to configure the cluster-level max_stream_duration for route-generated clusters, since the route-level max_stream_duration better reflects user intent.

If that’s not possible, then it would still be good to add a comment around the cluster-level max_stream_duration to explain what EG is doing here, and clarify that the cluster-level setting has no effect because the same value is configured at both levels.

@zirain zirain changed the title fix: MaxStreamDuration should worked on commonHttpProtocolOptions fix: MaxStreamDuration should worked for non-route cluster May 28, 2026
@zirain

zirain commented May 28, 2026

Copy link
Copy Markdown
Member Author

Updated to work for non-route cluster only.

Comment thread internal/xds/translator/cluster.go
@zhaohuabing

Copy link
Copy Markdown
Member

several non-route cluster callsites are not flagged. nonRouteCluster: true is only set in accesslog.go, tracing.go, and oidc.go (token
endpoint). These other constructors also build non-route clusters and accept a traffic parameter, so MaxStreamDuration will still be silently dropped:

  • createExtServiceXDSCluster (internal/xds/translator/utils.go:156) — used by ExtAuth, ExtProc, JWT, and the OIDC provider backend (extauth.go:268,277,
    extproc.go:185, jwt.go:326, oidc.go:422).
  • addClusterFromURL (internal/xds/translator/utils.go:190) — used by JWT JWKS URI (jwt.go:335) and DynamicModule remote (dynamicmodule.go:182)

Comment thread internal/xds/translator/cluster.go Outdated
unstructuredRefs []*unstructured.Unstructured
extensionMgr *extensionTypes.Manager
logger logging.Logger
nonRouteCluster bool

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Non-blocking nit: Can we flip the flag name to isRoute? It sounds more natural.

@zirain zirain force-pushed the fix/maxStreamDuration branch from 4ed5d66 to 901599b Compare May 29, 2026 04:47
- host: "1.2.3.4"
port: 50000
name: "direct-route-dest/backend/0"
traffic:

Copy link
Copy Markdown
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this's a route cluster, which result only maxConnectionDuration to be setted.

@arkodg arkodg left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM thanks

@zhaohuabing zhaohuabing left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM. Thanks!

@zhaohuabing zhaohuabing merged commit 0edf51a into envoyproxy:main May 29, 2026
85 of 91 checks passed
@zirain zirain deleted the fix/maxStreamDuration branch June 1, 2026 02:32
jukie added a commit that referenced this pull request Jun 5, 2026
* fix: restore last transition time in merge status conditions (#8962)

* fix: restore last transition time in merge status conditions

Signed-off-by: Rudrakh Panigrahi <[email protected]>

* add release note

Signed-off-by: Rudrakh Panigrahi <[email protected]>

---------

Signed-off-by: Rudrakh Panigrahi <[email protected]>
(cherry picked from commit 9430ab9)
Signed-off-by: jukie <[email protected]>

* fix: applyBackendTLSSetting panic (#8998)

* fix: applyBackendTLSSetting panic

Signed-off-by: zirain <[email protected]>
(cherry picked from commit b754b68)
Signed-off-by: jukie <[email protected]>

* fix(xds): hot-reload xDS server cert in GatewayNamespaceMode (#8959)

* fix(xds): hot-reload xDS server cert in GatewayNamespaceMode

Signed-off-by: nguyenptk <[email protected]>

* extract loadServerTLSConfig helper and add tests

Signed-off-by: nguyenptk <[email protected]>

---------

Signed-off-by: nguyenptk <[email protected]>
Signed-off-by: Nguyên (Harry) <[email protected]>
(cherry picked from commit 35c5a12)
Signed-off-by: jukie <[email protected]>

* fix(helm): correct typo in HPA maxReplicas required message (#9032)

Signed-off-by: benjaminch <[email protected]>
(cherry picked from commit 8da6dc7)
Signed-off-by: jukie <[email protected]>

* fix: allow IANA cipher names (#9073)

allow IANA cipher names

Signed-off-by: Huabing Zhao <[email protected]>
(cherry picked from commit ea41d6f)
Signed-off-by: jukie <[email protected]>

* fix(gatewayapi): prefer sectioned BackendTLSPolicy over wildcard for same target (#9068)

* add passing testcase showing wildcard overriding sectioned backendTLSPolicy

Signed-off-by: AlecDiraimondo <[email protected]>

* fix(gatewayapi): prefer sectioned BackendTLSPolicy over wildcard for same target

When two BackendTLSPolicies target the same backend — one with sectionName
set, one without — the wildcard wins whenever it sorts earlier, since
policies are iterated in creationTimestamp order with no specificity
preference. The sectioned policy is silently dropped: never selected by
the translator, no status written.

Make getBackendTLSPolicy prefer policies with an explicit sectionName
match before falling back to wildcard matches. Matches the specificity
ordering SecurityPolicy already implements.

Refs #9035

Signed-off-by: AlecDiraimondo <[email protected]>

* update release notes

Signed-off-by: AlecDiraimondo <[email protected]>

* fixes from review: dedup helper, return early if backendTarget has no section specified

Signed-off-by: AlecDiraimondo <[email protected]>
Co-authored-by: Rudrakh Panigrahi <[email protected]>

---------

Signed-off-by: AlecDiraimondo <[email protected]>
Signed-off-by: Alec Diraimondo <[email protected]>
Co-authored-by: Rudrakh Panigrahi <[email protected]>
(cherry picked from commit 07ec95c)
Signed-off-by: jukie <[email protected]>

* fix: encode TLS secrets with canonical PEM formatting (#8744)

* exclude non-ready https

Signed-off-by: Karol Szwaj <[email protected]>

* prevent BoringSSL errors

Signed-off-by: Karol Szwaj <[email protected]>

* Add trailing newline to testdata PEMs

Signed-off-by: Karol Szwaj <[email protected]>

* add release notes

Signed-off-by: Karol Szwaj <[email protected]>

* use shallow copy instead of DeepCopy

Signed-off-by: Karol Szwaj <[email protected]>

* use maps.Clone for the data and split the PRs

Signed-off-by: Karol Szwaj <[email protected]>

---------

Signed-off-by: Karol Szwaj <[email protected]>
(cherry picked from commit 560a262)
Signed-off-by: jukie <[email protected]>

* fix: MaxStreamDuration should worked for non-route cluster (#9028)

* fix: MaxStreamDuration should worked on commonHttpProtocolOptions

Signed-off-by: zirain <[email protected]>

* should only worked for non-route cluster

Signed-off-by: zirain <[email protected]>

* fix

Signed-off-by: zirain <[email protected]>

---------

Signed-off-by: zirain <[email protected]>
(cherry picked from commit 0edf51a)
Signed-off-by: jukie <[email protected]>

* fix(egctl): skip missing CRDs in `x status` bulk modes (#9099)

`egctl x status all` (and `xroute`/`xpolicy`) aborted with an error when a
Gateway API CRD was not installed in the cluster (e.g. TCPRoute). Missing
CRDs are now skipped, and reported on stderr when `-v` is set. Explicit
single-resource invocations still error loudly so users get a clear signal
when a kind they asked for is unavailable.

Signed-off-by: Jeremiah Snapp <[email protected]>
(cherry picked from commit 3ab110e)
Signed-off-by: jukie <[email protected]>

* fix(gatewayapi): reject unsupported BackendRef URLRewrite path rewrites (#9131)

gatewayapi: reject BackendRef URLRewrite path modifiers

Signed-off-by: Aditya7880900936 <[email protected]>
(cherry picked from commit acaf4cd)
Signed-off-by: jukie <[email protected]>

* fix: force HTTP1 for Service/ServiceImport with websocket appProtocol (#9138)

force HTTP1 for Service/ServiceImport with websocket appProtocol

Signed-off-by: Huabing Zhao <[email protected]>
(cherry picked from commit b1aa5d9)
Signed-off-by: jukie <[email protected]>

* fix: move validation admission policy outside of crds directory (#9024)

* move validation admission policy outside of crds directory

Signed-off-by: Huabing Zhao <[email protected]>

* fix upgrade

Signed-off-by: Huabing Zhao <[email protected]>

* add release note

Signed-off-by: Huabing Zhao <[email protected]>

* fix

Signed-off-by: Huabing Zhao <[email protected]>

* update docs

Signed-off-by: Huabing Zhao <[email protected]>

* address comment

Signed-off-by: Huabing Zhao <[email protected]>

* address comment

Signed-off-by: Huabing Zhao <[email protected]>

* address comments

Signed-off-by: Huabing Zhao <[email protected]>

* update relese note

Signed-off-by: Huabing Zhao <[email protected]>

* address comments

Signed-off-by: Huabing Zhao <[email protected]>

* fix gen

Signed-off-by: Huabing Zhao <[email protected]>

* use a more general name for this option

Signed-off-by: Huabing Zhao <[email protected]>

* update docs

Signed-off-by: Huabing Zhao <[email protected]>

* rename

Signed-off-by: Huabing Zhao <[email protected]>

* update test

Signed-off-by: Huabing Zhao <[email protected]>

* update release note

Signed-off-by: Huabing Zhao <[email protected]>

* fix gen

Signed-off-by: Huabing Zhao <[email protected]>

* rename supportingResources to safeUpgradePolicy

Signed-off-by: Huabing Zhao <[email protected]>

* update

Signed-off-by: Huabing Zhao <[email protected]>

---------

Signed-off-by: Huabing Zhao <[email protected]>
Signed-off-by: Huabing (Robin) Zhao <[email protected]>
Co-authored-by: zirain <[email protected]>
(cherry picked from commit 35ad8f2)
Signed-off-by: jukie <[email protected]>

* fix: alway include the controller namespace in the watched namespaces (#9100)

* always include the controller namespace in the k8s provider

Signed-off-by: Huabing Zhao <[email protected]>

* update docs

Signed-off-by: Huabing Zhao <[email protected]>
(cherry picked from commit 84be503d855e968b7b6999d476a6897e98d4e330)

* address comment

Signed-off-by: Huabing Zhao <[email protected]>

* address comments

Signed-off-by: Huabing Zhao <[email protected]>

* address comments

Signed-off-by: Huabing Zhao <[email protected]>

* Potential fix for pull request finding

Co-authored-by: Copilot Autofix powered by AI <[email protected]>
Signed-off-by: Huabing (Robin) Zhao <[email protected]>

* minor changes

Signed-off-by: Huabing Zhao <[email protected]>

* fix test

Signed-off-by: Huabing Zhao <[email protected]>

* fix lint

Signed-off-by: Huabing Zhao <[email protected]>

* fix test

Signed-off-by: Huabing Zhao <[email protected]>

* fix test

Signed-off-by: Huabing Zhao <[email protected]>

* fix test

Signed-off-by: Huabing Zhao <[email protected]>

---------

Signed-off-by: Huabing Zhao <[email protected]>
Signed-off-by: Huabing (Robin) Zhao <[email protected]>
Co-authored-by: Copilot Autofix powered by AI <[email protected]>
(cherry picked from commit 5266322)
Signed-off-by: jukie <[email protected]>

* test(helm): regenerate safe-upgrade-policy-disabled golden for release/v1.8

The safe-upgrade-policy-disabled test case added in #9024 shipped a golden
rendered against main's chart templates, which include features not present
on release/v1.8 (automountServiceAccountToken, expanded HPA RBAC verbs,
serviceaccount securityContext, ArgoCD/Flux hook comment). Regenerate the
golden against the release branch templates and restore the release-pinned
ratelimit image tag (ff287602).

Signed-off-by: jukie <[email protected]>

* regen

Signed-off-by: jukie <[email protected]>

* chore: bump golang to 1.26.4 (#9147)

Signed-off-by: zirain <[email protected]>
Signed-off-by: jukie <[email protected]>

---------

Signed-off-by: Rudrakh Panigrahi <[email protected]>
Signed-off-by: jukie <[email protected]>
Signed-off-by: zirain <[email protected]>
Signed-off-by: nguyenptk <[email protected]>
Signed-off-by: Nguyên (Harry) <[email protected]>
Signed-off-by: benjaminch <[email protected]>
Signed-off-by: Huabing Zhao <[email protected]>
Signed-off-by: AlecDiraimondo <[email protected]>
Signed-off-by: Alec Diraimondo <[email protected]>
Signed-off-by: Karol Szwaj <[email protected]>
Signed-off-by: Jeremiah Snapp <[email protected]>
Signed-off-by: Aditya7880900936 <[email protected]>
Signed-off-by: Huabing (Robin) Zhao <[email protected]>
Co-authored-by: Rudrakh Panigrahi <[email protected]>
Co-authored-by: zirain <[email protected]>
Co-authored-by: Nguyên (Harry) <[email protected]>
Co-authored-by: BenjaminCh <[email protected]>
Co-authored-by: Huabing (Robin) Zhao <[email protected]>
Co-authored-by: Alec Diraimondo <[email protected]>
Co-authored-by: Karol Szwaj <[email protected]>
Co-authored-by: Jeremiah Snapp <[email protected]>
Co-authored-by: Aditya Sanskar Srivastav <[email protected]>
Co-authored-by: Copilot Autofix powered by AI <[email protected]>
cleman95 pushed a commit to cleman95/gateway that referenced this pull request Jun 25, 2026
…y#9028)

* fix: MaxStreamDuration should worked on commonHttpProtocolOptions

Signed-off-by: zirain <[email protected]>

* should only worked for non-route cluster

Signed-off-by: zirain <[email protected]>

* fix

Signed-off-by: zirain <[email protected]>

---------

Signed-off-by: zirain <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants