HTTP Server Configuration

Configure the HTTP server that handles SSE connections.

Configuration Options

bind_address

The address and port to listen on.

[http]
bind_address = "0.0.0.0:8080"

Value	Description
`0.0.0.0:8080`	Listen on all interfaces, port 8080
`127.0.0.1:8080`	Listen on localhost only
`192.168.1.10:8080`	Listen on specific interface

max_connections

Maximum number of concurrent client connections.

[http]
max_connections = 100000

When the limit is reached, new connections receive 503 Service Unavailable.

Considerations: - Each connection uses ~2KB of memory - Set based on available RAM and file descriptors - Monitor connections_active metric

connection_timeout_sec

Time in seconds before idle connections are closed.

[http]
connection_timeout_sec = 300

Recommendations: - Mobile clients: 60-120 seconds (battery considerations) - Web clients: 300-600 seconds - Internal services: 3600+ seconds

keepalive_interval_sec

Interval for sending SSE keepalive comments.

[http]
keepalive_interval_sec = 30

Keepalives prevent: - Proxy/load balancer timeouts - Client connection detection issues - NAT table expiration

Format sent to clients:

: keepalive

max_message_queue_size

Maximum pending messages per connection.

[http]
max_message_queue_size = 1000

When the queue is full, backpressure_policy determines behavior.

thread_pool_size

Number of worker threads for handling connections.

[http]
thread_pool_size = 0  # Auto-detect CPU count

Value	Behavior
`0`	Auto-detect (recommended)
`N`	Use exactly N threads

Guidelines: - CPU-bound: Match CPU core count - I/O-bound: 2x CPU core count - Memory-constrained: Reduce threads

Request Size Limits

Protect against malicious requests:

[http]
max_request_line_size = 8192   # Maximum URL length
max_header_size = 8192         # Maximum single header size
max_headers = 100              # Maximum number of headers

Requests exceeding limits receive 413 Request Entity Too Large.

Rate Limiting

Per-IP rate limiting configuration:

[http]
rate_limit_per_ip = 100      # Requests per window
rate_limit_window_sec = 60   # Window duration in seconds

Example: 100 requests per 60 seconds = ~1.67 requests/second average

Clients exceeding the limit receive 429 Too Many Requests.

Tip

Set higher limits for trusted internal networks using a reverse proxy.

cors_allowed_origins

CORS (Cross-Origin Resource Sharing) configuration:

[http]
cors_allowed_origins = "*"                    # Allow all origins
# cors_allowed_origins = "https://app.example.com"  # Specific origin

Response headers set:

Access-Control-Allow-Origin: <configured-value>
Access-Control-Allow-Methods: GET, OPTIONS
Access-Control-Allow-Headers: Authorization, Content-Type

backpressure_policy

Behavior when message queue is full:

[http]
backpressure_policy = "drop_newest"

Policy	Behavior	Use Case
`drop_newest`	Reject new messages	Prefer historical data
`drop_oldest`	Remove oldest messages	Prefer latest data
`block`	Wait for queue space	Guaranteed delivery

Performance Tuning

High Connection Count

For 100k+ connections:

[http]
bind_address = "0.0.0.0:8080"
max_connections = 500000
thread_pool_size = 32
max_message_queue_size = 100
keepalive_interval_sec = 60

System tuning required:

# Increase file descriptor limit
ulimit -n 1000000

# Kernel parameters
sysctl -w net.core.somaxconn=65535
sysctl -w net.ipv4.tcp_max_syn_backlog=65535

Low Latency

For minimal message latency:

[http]
thread_pool_size = 0           # Match CPU cores
max_message_queue_size = 10    # Small queue
backpressure_policy = "drop_oldest"

Memory Constrained

For limited memory environments:

[http]
max_connections = 10000
thread_pool_size = 4
max_message_queue_size = 50

Example Configurations

Development

[http]
bind_address = "127.0.0.1:8080"
max_connections = 100
connection_timeout_sec = 3600
keepalive_interval_sec = 60
rate_limit_per_ip = 1000
cors_allowed_origins = "*"

Production

[http]
bind_address = "0.0.0.0:8080"
max_connections = 100000
connection_timeout_sec = 300
keepalive_interval_sec = 30
max_message_queue_size = 500
thread_pool_size = 0
max_request_line_size = 4096
max_header_size = 4096
max_headers = 50
rate_limit_per_ip = 100
rate_limit_window_sec = 60
cors_allowed_origins = "https://app.example.com"
backpressure_policy = "drop_newest"

Monitoring

Key metrics to monitor:

Metric	Description	Alert Threshold
`connections_active`	Current connections	> 80% of max
`connections_rejected`	Rejected connections	> 0 sustained
`queue_depth`	Message queue size	> 80% of max
`rate_limit_hits`	Rate limit triggers	Unusual spikes