Appwrite SDK‑Server Concurrency Limit: Configurable via Environment Variable?

Answer

Appwrite does not expose an environment variable named APPWRITE_MAX_CONCURRENCY. The server runs on PHP with Swoole (via the Utopia framework), not Go. Concurrent request capacity is determined by the number of Swoole worker processes, which is derived from CPU cores and a per-core multiplier.

Environment Variables That Affect Concurrency

• _APP_WORKER_PER_CORE (default: 4) – multiplier applied to the number of CPU cores to set the target worker count.
• _APP_MAX_WORKERS (default: 0, meaning unlimited) – hard ceiling on worker processes.
• _APP_MIN_WORKERS (default: 0) – minimum workers to keep alive.

The effective worker count at startup is min(max(cpu_cores * _APP_WORKER_PER_CORE, _APP_MIN_WORKERS), _APP_MAX_WORKERS). Each worker can handle many concurrent connections asynchronously via Swoole's event loop, so the practical concurrency limit is workers × connections-per-worker, bounded by OS file-descriptor limits and memory.

Where the Limit Is Defined in Source

In the Appwrite repository (appwrite/appwrite), the Swoole HTTP server is bootstrapped in src/Appwrite/App.php (method runHttp). The worker count is calculated in src/Appwrite/Config.php or the Utopia framework's Swoole wrapper. Search for _APP_WORKER_PER_CORE in the codebase to trace the exact logic for your version.

Verifying the Effective Limit in a Running Instance

1. Check configured worker count from environment

   docker exec  env | grep -E '_APP_WORKER|_APP_MAX_WORKERS|_APP_MIN_WORKERS'
   

2. Observe actual Swoole worker processes

   docker exec  ps aux | grep 'swoole' | grep -v grep
   

   Count the worker lines (exclude master/manager processes). This reflects the active concurrency slots.
3. Query Swoole server stats (if stats endpoint enabled) Some deployments expose /v1/health or a Swoole status page. If enabled, the response includes worker_num, idle_workers, and connection counts.
4. Load test and correlate latency

   wrk -t4 -c200 -d30s http://host/v1/health
   

   Watch for the latency inflection point; it typically aligns with worker_num × Swoole's max_connections_per_worker (default ~10,000) or the OS ulimit -n.

What to Adjust If You Need More Concurrency

• Increase _APP_WORKER_PER_CORE (e.g., to 8) or set _APP_MAX_WORKERS explicitly.
• Raise the container/file-descriptor limit: docker run --ulimit nofile=65536:65536 ….
• Scale horizontally: run multiple Appwrite containers behind a load balancer (each manages its own worker pool).

Missing Diagnostic Detail

Please share your Appwrite version (e.g., 1.5.2, 1.6.0) and whether you run the official Docker image or a custom build. The exact variable names and source locations shifted slightly between 1.4.x and 1.5.x, so the version determines which config keys are authoritative.