[moutyque]

Problem

In clustered mode, the leader can crash with:

fatal error: concurrent map iteration and map write

goroutine ... [running]:
github.com/openconfig/gnmic/pkg/app.(*App).dispatchTargetsOnce(...)
        pkg/app/clustering.go:288
github.com/openconfig/gnmic/pkg/app.(*App).dispatchTargets(...)
        pkg/app/clustering.go:273
created by github.com/openconfig/gnmic/pkg/app.(*App).startCluster.func1
        pkg/app/clustering.go:161

dispatchTargetsOnce ranges over a.Config.Targets while holding only dispatchLock.
The loader writes the same map under configLock (startLoader, a.Config.Targets[add.Name] = add).
Since the loader releases configLock before it dispatches (v0.49.1), it writes to the map more often while the dispatch loop runs.

We hit it in production on v0.49.1 with a 6-instance cluster and about 6000 targets from the HTTP loader: the leader crashed during the initial load after a rollout, leadership moved to another instance, which crashed the same way. It settled once the loader stopped adding targets.

Fix

Copy the targets under configLock.RLock() and iterate over the copy. The lock is held only for the copy, not across dispatchTarget, so the loader and the API handlers are never blocked by a dispatch pass.

Test

TestDispatchTargetsOnceSafeAgainstConcurrentLoaderWrites runs dispatchTargetsOnce while another goroutine adds targets under configLock.

• Before the fix: go test -race reports DATA RACE and the run dies with fatal error: concurrent map iteration and map write.
• After the fix: go test -race ./pkg/app/ passes.

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.