a

[BugsGroundhog]

Note

Medium Risk
Medium risk due to broad dependency upgrades (Go 1.15, etcd/nats/protobuf/validator) and behavior changes in cluster service discovery/watchers and pipeline handler signatures that can affect runtime concurrency and request processing.

Overview
Switches CI from Travis to GitHub Actions (tests.yaml), updates README build badge, and bumps the project to Go 1.15 with large dependency upgrades (notably go.etcd.io/etcd, NATS v2, protobuf, validator v10).

Improves cluster/runtime behavior: etcd service discovery is refactored to explicitly (re)initialize the etcd client, better handle dead lease/watch channels, reduce initial sync round-trips, and avoid watcher/sync races; gRPC server now binds to a configurable address; RPC tracing span names are clarified.

Updates request/pipeline and messaging internals: before-pipeline/struct validator signatures now return (context, out, err) to allow context propagation, request IDs switch to nuid, message route dictionaries become thread-safe, NATS RPC server drains on shutdown, and several tests/e2e/demo code are adjusted accordingly (including regenerated demo protos and new protos-compile-demo target).

^{Reviewed by Cursor Bugbot for commit 07f2cd1. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 4 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 07f2cd1. Configure here.}

[cursor]

Unbuffered channel causes potential deadlock during Init

High Severity

syncServersRunning is an unbuffered channel (line 93). In Init(), go sd.watchEtcdChanges() is launched, then sd.bootstrap() is immediately called, which calls SyncServers(true). SyncServers blocks on sd.syncServersRunning <- true until the watcher goroutine's inner select is ready to receive. Since watchEtcdChanges must first call sd.cli.Watch() and then launch its own inner goroutine before reaching select, there's a race where the main goroutine blocks before any receiver exists — a potential deadlock.

Additional Locations (1)

• cluster/etcd_service_discovery.go#L386-L391

 

^{Reviewed by Cursor Bugbot for commit 07f2cd1. Configure here.}

[cursor]

Server added locally despite etcd Put failure

Medium Severity

addServerIntoEtcd unconditionally calls sd.addServer(server) on line 221 after the cli.Put() call, regardless of whether the Put succeeded or failed. If the etcd write fails, the server is added to local maps but not persisted in etcd, leaving the service discovery in an inconsistent state. The caller (bootstrapServer) checks the returned error, but by then addServer has already modified local state.

 

^{Reviewed by Cursor Bugbot for commit 07f2cd1. Configure here.}

[cursor]

Watcher reconnect counter reset delays recovery unnecessarily

Medium Severity

When InitETCDClient() fails at line 669, failedWatchAttempts is reset to 0 at line 670. Since the watch channel chn is still closed, every subsequent select iteration immediately hits the !ok branch, but now must accumulate 10 more failures before retrying InitETCDClient. This unnecessarily delays reconnection attempts by ~10 seconds on each failed reconnect, rather than retrying immediately on the next iteration.

 

^{Reviewed by Cursor Bugbot for commit 07f2cd1. Configure here.}

[cursor]

Mock recorder missing argument for updated SyncServers signature

Medium Severity

The MockServiceDiscovery.SyncServers method was updated to accept a firstSync bool parameter, but the corresponding MockServiceDiscoveryMockRecorder.SyncServers() recorder method was not updated to accept a matching argument. This prevents tests from setting expectations with specific firstSync values and may cause gomock argument-count mismatches at runtime.

 

^{Reviewed by Cursor Bugbot for commit 07f2cd1. Configure here.}