fix(watch): do not drop events when a read carries multiple responses - #224
Conversation
read_watch() splits one body read into several watch responses, decodes the events of each of them into all_events, and then attaches all_events to `body`, which points at the last decoded response. The attach was guarded by `#all_events > 1`, so a read holding [response with 1 event, response without events] returned the last response as is and the event was lost, silently and unrecoverably for the caller. Today etcd never produces that layout because the only event-less responses are `created` and `canceled`, which never trail an event response. It does as soon as the watch is created with progress_notify, which apache/apisix needs to advance start_revision safely (apache/apisix#13067). Attach the events whenever any were collected. Also guard on body.result, which is nil when the last response is an error body with an http code below 500 - reachable before this change too, and a nil index in the watch loop.
|
Warning Review limit reachedYou’ve reached a temporary PR review limit under our Fair Usage Limits Policy. Next review available in: 9 minutes Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available. How can I continue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews. How do review limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please refer docs for additional details. 📝 WalkthroughWalkthroughThe v3 watch response merger now preserves events from coalesced responses. A regression test covers an event response followed by a progress response, the TLS test accepts alternate certificate wording, and the CI build environment updates its runner and Go tooling. ChangesWatch response coalescing
CI and TLS test maintenance
Estimated code review effort: 2 (Simple) | ~10 minutes Sequence Diagram(s)sequenceDiagram
participant WatchTest
participant FakeNginx
participant EtcdClient
participant WatchMerger
WatchTest->>EtcdClient: call watchdir("/test")
EtcdClient->>FakeNginx: request /v3/watch
FakeNginx-->>EtcdClient: concatenated event and progress JSON fragments
EtcdClient->>WatchMerger: decode and merge watch responses
WatchMerger-->>WatchTest: revision 9 with one preserved event
🚥 Pre-merge checks | ✅ 6✅ Passed checks (6 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
The ubuntu-20.04 hosted runner image has been retired, so jobs pinned to it queue indefinitely with no runner to pick them up. Move to ubuntu-22.04.
goreman's transitive dependency golang.org/x/sys v0.46.0 now imports the stdlib "slices" package (Go >= 1.21), so building it under the pinned Go 1.17 fails at the install step. Bump the toolchain to 1.22. Since 'go get' no longer installs binaries on Go >= 1.18, switch to 'go install github.com/mattn/goreman@latest'.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In @.github/workflows/ci.yml:
- Around line 38-42: Update the goreman installation step in the workflow to pin
github.com/mattn/goreman to a release compatible with Go 1.22, or raise the
setup-go version to match the module’s declared requirement. Keep the setup-go
and goreman versions consistent so installation does not trigger an unintended
toolchain download.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 208b83e0-07cf-48cd-911b-7e5b2b3c34d5
📒 Files selected for processing (1)
.github/workflows/ci.yml
There was a problem hiding this comment.
Pull request overview
Fixes a v3 watch edge case where multiple newline-delimited watch responses arrive in a single HTTP body read and events from earlier responses could be silently lost when the last response is event-less (e.g., progress notifications).
Changes:
- Preserve collected watch events by attaching them whenever any events were decoded (and avoid indexing
nilwhen the last response is an error body withoutresult). - Add a Test::Nginx case that simulates an event response immediately followed by a progress notification in the same body chunk.
- Refresh CI environment (Ubuntu 22.04, newer
setup-go, and goreman installation method).
Reviewed changes
Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.
| File | Description |
|---|---|
lib/resty/etcd/v3.lua |
Ensures watch events aren’t dropped when a single read contains multiple watch responses and the last response has no events. |
t/v3/watch.t |
Adds regression coverage for coalesced watch responses (event + progress notify in one read). |
.github/workflows/ci.yml |
Updates CI runner/tooling and switches goreman installation to go install. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
The new t/v3/watch.t TEST 1 had no '--- request' section, so Test::Nginx had no request to send to the /t location and bailed out with 'Request line should be non-empty' before the assertions ran.
OpenSSL 3.x reworded the X509 verify error from 'self signed certificate' to 'self-signed certificate' (hyphen). Broaden the TEST 2 pattern to 'self[ -]signed' so it matches both the 1.1.x and 3.x spellings.
Addresses review feedback (CodeRabbit/Copilot): 'go install goreman@latest' is non-deterministic, and goreman >= v0.3.19 declares 'go 1.25' in its module, so Go's default GOTOOLCHAIN=auto would silently fetch a 1.25 toolchain and defeat the setup-go 1.22 pin. Pin to v0.3.15 (module 'go 1.19', golang.org/x/sys v0.6.0) which builds with the 1.22 toolchain, and set GOTOOLCHAIN=local so a future pin bump that needs a newer toolchain fails loudly instead of downloading one.
The fix's premise is that the response holding events is not necessarily the last one in a coalesced read, so cover both orderings and an interleaving, not just [event, progress]: - TEST 2: [progress, event] -> event is the last response - TEST 3: [event, progress, event] -> both events survive, in order Also document why TEST 1 reports the trailing progress notification's revision while still delivering the earlier event.
1.10.7 ships the read_watch coalescing fix (api7/lua-resty-etcd#224) that progress_notify depends on: when etcd coalesces an event response and a progress notification into a single read, events are no longer dropped.
Problem
read_watch()splits one body read into several watch responses, decodes the events of each intoall_events, and then attachesall_eventstobody— which points at the last decoded response. The attach was guarded by#all_events > 1:So a read holding
[response with 1 event, response without events]returns the last response as is, and the event is silently lost. The caller cannot detect it.Today etcd never produces that layout, because the only event-less responses are
createdandcanceledand neither trails an event response. It does as soon as the watch is created withprogress_notify: a notification can be coalesced right behind an event response on the same TCP read.apache/apisix needs
progress_notifyto advancestart_revisionsafely (apache/apisix#13067), so this has to be fixed first — otherwise the same class of silent event loss is reintroduced inside the library.Fix
Attach the events whenever any were collected (
> 0).Also guard on
body.result, which isnilwhen the last response is an error body with an http code below 500. That path is reachable today too and> 1would already indexnilthere, so this is a strict improvement rather than a new behaviour.Compatibility
No behaviour change for anything that exists today:
#chunks == 1returns early for event-less responses)[notify, event response]order:bodyalready is the event response, so the assignment is a no-op[event, canceled]: both spellings end in the samecompactedfull-reload path on the apisix sideTest
t/v3/watch.tdrives the client against a fake watch endpoint that writes an event response followed by a progress notification in a single body chunk, and asserts the event survives. A real etcd cannot be pushed into producing that coalescing reliably.Not run locally — CI covers it.
Follow-up
A
v1.10.7release is needed before apache/apisix can pin it and merge its side.Summary by CodeRabbit
Bug Fixes
Tests
Chores