feat: Add /health endpoint for server monitoring #3460
Conversation
- Add configurable health endpoint to Dash core - Support custom endpoint paths and disable option - Include system metrics and callback information - Add comprehensive test suite with 10 test cases - Compatible with load balancers, Docker, Kubernetes Resolves: Add health check endpoint to Dash framework
lfelipediniz
requested review from
T4rk1n,
ndrezn,
gvwilson and
emilykl
as code owners
gvwilson
added
feature
|
@T4rk1n please let us know what you think - thanks |
|
cc @selfeki |
dash/dash.py
Outdated
| def serve_health(self): | ||
| """ | ||
| Health check endpoint for monitoring Dash server status. | ||
| Returns a JSON response indicating the server is running and healthy. | ||
| This endpoint can be used by load balancers, monitoring systems, | ||
| and other platforms to check if the Dash server is operational. | ||
| :return: JSON response with status information | ||
| """ | ||
| import datetime | ||
| import platform | ||
| import psutil |
There was a problem hiding this comment.
I don't think there should be any code in the default handler, it should simply returns "OK" with response 200. The api should provide a way for the developer to add a function to handle the response and do it's health check properly by checking the needed services.
There was a problem hiding this comment.
Agree with @T4rk1n, I would expect the /health response to be as simple as
{"status":"ok"}
The remaining metadata is still useful if we return it via a different endpoint, eg. /info
dash/dash.py
Outdated
| def serve_health(self): | ||
| """ | ||
| Health check endpoint for monitoring Dash server status. | ||
| Returns a JSON response indicating the server is running and healthy. | ||
| This endpoint can be used by load balancers, monitoring systems, | ||
| and other platforms to check if the Dash server is operational. | ||
| :return: JSON response with status information | ||
| """ | ||
| import datetime | ||
| import platform | ||
| import psutil | ||
| import sys | ||
|
|
||
| # Basic health information | ||
| health_data = { | ||
| "status": "healthy", | ||
| "timestamp": datetime.datetime.utcnow().isoformat() + "Z", | ||
| "dash_version": __version__, | ||
| "python_version": sys.version, | ||
| "platform": platform.platform(), | ||
| } | ||
|
|
||
| # Add server information if available | ||
| try: | ||
| health_data.update({ | ||
| "server_name": self.server.name, | ||
| "debug_mode": self.server.debug, | ||
| "host": getattr(self.server, 'host', 'unknown'), | ||
| "port": getattr(self.server, 'port', 'unknown'), | ||
| }) | ||
| except Exception: | ||
| pass | ||
|
|
||
| # Add system resource information if psutil is available | ||
| try: | ||
| health_data.update({ | ||
| "system": { | ||
| "cpu_percent": psutil.cpu_percent(interval=0.1), | ||
| "memory_percent": psutil.virtual_memory().percent, | ||
| "disk_percent": psutil.disk_usage('/').percent if os.name != 'nt' else psutil.disk_usage('C:').percent, | ||
| } | ||
| }) | ||
| except ImportError: | ||
| # psutil not available, skip system metrics | ||
| pass | ||
| except Exception: | ||
| # Error getting system metrics, skip them | ||
| pass | ||
|
|
||
| # Add callback information | ||
| try: | ||
| health_data.update({ | ||
| "callbacks": { | ||
| "total_callbacks": len(self.callback_map), | ||
| "background_callbacks": len(getattr(self, '_background_callback_map', {})), | ||
| } | ||
| }) | ||
| except Exception: | ||
| pass | ||
|
|
||
| return flask.jsonify(health_data) |
There was a problem hiding this comment.
Normal health check should return "OK", 200
There was a problem hiding this comment.
This is a great idea! I'll let those more familiar with the codebase give a detailed view.
I'd like to see a /ready endpoint as well. Can one be added as part of this work? It should be possible to reuse a lot of this code. To summarize the difference:
- Health checks ("is this container responding at all?") is relevant to the cluster scheduler.
- Readiness checks ("is this container able to respond to this class of requests right now?") are relevant to the load balancer.
For example, if a service has an external database connection, one might want to have the health check just make sure the app can respond to traffic, but have the readiness check test that the DB is reachable and can perform a basic query.
|
Thanks for the analysis, guys! So, after reading what you sent, I believe that for this issue focused on the
Regarding Does this plan work for ? |
|
Looks 👍 , but I would recommend removing the system metadata from this PR and creating a second PR for that feature; feels like there's some more discussion to be had on whether that should be in a new endpoint that I don't think is worth blocking the initial implementation for. |
- Add optional health_endpoint parameter to Dash constructor (default: None) - Implement simple health check endpoint returning 'OK' with HTTP 200 - Health endpoint respects routes_pathname_prefix configuration - Add comprehensive unit tests for health endpoint functionality - Remove integration tests in favor of simpler unit tests - Apply code formatting with black
🚑 Add Health Check Endpoint to Dash Core (#3446)
This PR introduces a built-in
/healthendpoint for Dash applications, designed to support load balancers, Docker health checks, Kubernetes probes, monitoring scripts, and CI/CD pipelines.✨ Key Features
Adds
/healthendpoint to all Dash appsConfigurable path support (respects
routes_pathname_prefix)Toggle to disable endpoint if needed
Includes detailed metadata:
psutil)Includes comprehensive test suite with 10 test cases
Endpoint is unauthenticated by default to support external probes
✅ Example Output
{ "status": "healthy", "timestamp": "2025-09-30T00:35:03.982409Z", "dash_version": "3.2.0", "python_version": "3.12.3", "platform": "Linux-6.8.0-71-generic-x86_64", "server_name": "my_dash_app", "debug_mode": true, "host": "127.0.0.1", "port": "8050", "system": { "cpu_percent": 17.8, "memory_percent": 37.9, "disk_percent": 52.6 }, "callbacks": { "total_callbacks": 4, "background_callbacks": 0 } }🔧 Configuration Notes
routes_pathname_prefixpsutilis optional - system metrics are included only if available🔬 Testing
A full suite of 10 tests has been added, covering:
psutilAll tests pass locally :)