Firebase函数跨平台部署及Firestore变更监听技术咨询

Hey there, let's tackle your two questions head-on—I’ve dealt with similar Firebase Spark plan constraints before, so I get where you’re coming from!

1. Can you deploy Firestore change-triggered functions to non-Google platforms like Zeit/AWS Lambda? #

Short answer: You can’t directly deploy official firebase-functions triggers (like onCreate/onUpdate) to non-Google cloud platforms. Those triggers are tightly tied to Google Cloud Functions (GCF) infrastructure—they rely on GCF’s native event routing system that’s integrated exclusively with Firebase services.

But here’s a solid workaround: Use the Firebase Admin SDK (which works in any Node.js/Java/Python environment) to build your own Firestore change listener in AWS Lambda, Vercel (formerly Zeit), or any other serverless platform. Here’s how it works:

• Skip GCF’s managed triggers entirely and use Firestore’s Change Streams (via the Admin SDK) to subscribe to document create/update/delete events. Change Streams let you track changes across an entire collection, and you can persist a cursor to resume listening if your serverless function restarts.
• Watch out for serverless platform limits: Lambda functions have maximum execution time caps (e.g., 15 minutes for AWS), so you’ll need to handle long-running connections carefully. A common approach is to save your last known stream position to a persistent store (like Firestore itself), then have the function exit and restart later to pick up where it left off.
• You’ll need to authenticate the Admin SDK with a Firebase service account key (downloadable from the Firebase Console) in your non-Google environment—just make sure to store this key securely (use environment variables, never hardcode it).

2. Can you use onSnapshot instead of onCreate/onUpdate to listen for Firestore changes? #

Absolutely, but there are key differences to keep in mind based on your use case:

• onSnapshot is a real-time listener available in both the Firebase Client SDK and Admin SDK. When you initialize it, it first sends a full snapshot of the documents/collection you’re targeting, then triggers updates whenever a document is created, modified, or deleted.
• Compare that to firebase-functions triggers like onCreate: Those are event-driven, fire only once per matching action (no initial snapshot), and are fully managed by GCF.
• For your full-text search workaround, onSnapshot could work in a server-side environment (like your non-Google Lambda), but again, be mindful of serverless execution limits. A better fit for server-side use might be Firestore Change Streams—they’re designed for reliable, long-running change tracking and support resumable cursors, which onSnapshot doesn’t handle natively if your connection drops or function restarts.
• If you’re running the listener in a long-running service (not serverless), onSnapshot is totally reliable for keeping your search index in sync with Firestore.

A quick side note for your full-text search goal: Since you can’t use Algolia, you might consider building a simple inverted index directly in Firestore (storing keywords as fields in a dedicated collection) or checking if Google’s Cloud Search is allowed in the Spark plan. But that’s just extra context—focused on your original questions!

内容的提问来源于stack exchange，提问作者Misha Moroshko