boost-bootstrap-adapter-nats

REQUIRED BACKGROUND:

• boost-bootstrap-function — handler typing rule.
• boost-bootstrap-middleware — recovery/logger/publisher chain.
• boost-extra-middleware — NewAnyErrorWrapper for the workaround.
• boost-bootstrap-adapter-pubsub — same shape, full workaround pattern documented there.

Canonical (prototype / dev)

import (
    anats "github.com/xgodev/boost/bootstrap/function/adapter/contrib/nats-io/nats.go/v1"
    "github.com/xgodev/boost/bootstrap/function"
)

fn, _ := function.New[*cloudevents.Event](rec, lmi, pmi)
fn.Run(ctx, handle, anats.New[*cloudevents.Event](conn))

Subscriptions are configured via boost.bootstrap.function.adapter.nats.subjects (comma-separated) and boost.bootstrap.function.adapter.nats.queueGroup. Override at deploy via BOOST_BOOTSTRAP_FUNCTION_ADAPTER_NATS_*.

Production caveat — same ctx-loss as Pub/Sub

bootstrap/function/adapter/contrib/nats-io/nats.go/v1/helper.go:44/46 and subscriber.go:62 hard-code context.Background(). SIGTERM does not gracefully drain.

Apply the same workaround pattern documented in boost-bootstrap-adapter-pubsub: bypass fn.Run, build the chain via extra/middleware.NewAnyErrorWrapper, drive anats.NewSubscriber with a signal-aware ctx, and add the // TODO(boost-upstream): annotation naming the offending file.

Queue groups

NATS queue groups distribute messages across N subscribers (load balancing). Configure via boost.bootstrap.function.adapter.nats.queueGroup. Without a queue group, every subscriber gets every message (broadcast).

Red flags

Red flag	Fix
nats.Conn.Subscribe(...) directly from the upstream SDK	Use anats.NewSubscriber(...).Subscribe(ctx) or function.New + fn.Run
Bypass of fn.Run without // TODO(boost-upstream): naming helper.go:44/subscriber.go:62	Add the comment, OR accept ungraceful shutdown
Multiple NATS connections per process	Construct one *nats.Conn at startup, share
Config tunables read via os.Getenv	Use BOOST_BOOTSTRAP_FUNCTION_ADAPTER_NATS_* overrides