APISIX Configuration Reference

APISIX configuration lives in layers/api-gateway/config/apisix.yaml. This file is mounted into the APISIX container and controls the core proxy, plugin loading, admin API, etcd connection, and observability.

Core Settings

apisix:
  node_listen: 9080
  enable_admin: true
  enable_websocket: true
  admin_listen:
    port: 9180
  extra_lua_path: "/usr/local/apisix/apisix/plugins/custom/?.lua;/usr/local/apisix/apisix/plugins/custom/?/init.lua;"

Setting	Value	Description
node_listen	9080	Main proxy port. All API traffic enters here.
enable_admin	true	Enables the Admin API for route/consumer management.
enable_websocket	true	Enables WebSocket proxying.
admin_listen.port	9180	Admin API port.
extra_lua_path	(see above)	Adds the custom plugin directory to the Lua module search path.

Deployment

Admin API

deployment:
  admin:
    admin_key:
      - name: admin
        key: $\{\{APISIX_ADMIN_KEY\}\}
        role: admin
    allow_admin:
      - 127.0.0.0/24
      - 172.16.0.0/12
      - 192.168.0.0/16
      - 10.0.0.0/8

Setting	Description
admin_key	Admin API key. Loaded from APISIX_ADMIN_KEY environment variable. Never hardcode.
allow_admin	CIDR blocks allowed to access the Admin API. Restricted to private networks.

DANGER

The Admin API key is set via the $\{\{APISIX_ADMIN_KEY\}\} environment variable. Never commit this value. Use the X-API-KEY header for all Admin API requests.

Example Admin API call:

curl http://localhost:9180/apisix/admin/routes \
  -H "X-API-KEY: $APISIX_ADMIN_KEY"

etcd Connection

deployment:
  etcd:
    host:
      - "http://etcd:2379"
    prefix: "/apisix"
    timeout: 30

Setting	Value	Description
host	http://etcd:2379	etcd endpoint for APISIX configuration storage.
prefix	/apisix	Key prefix in etcd. All APISIX config stored under this path.
timeout	30	Connection timeout in seconds.

Plugins

All enabled plugins are declared in the plugins list. Only plugins listed here can be used in route configurations.

Standard Plugins

Rate Limiting

Plugin	Description
limit-count	Fixed window counter rate limiting
limit-req	Request rate limiting (leaky bucket)
limit-conn	Concurrent connection limiting

Authentication

Plugin	Description
jwt-auth	JWT validation (used with Keycloak)
key-auth	API key authentication
basic-auth	HTTP Basic authentication
openid-connect	OpenID Connect (Keycloak integration)

Traffic Management

Plugin	Description
proxy-rewrite	Rewrite upstream URI, host, headers
response-rewrite	Rewrite response headers/body
redirect	HTTP redirect
cors	Cross-Origin Resource Sharing
ip-restriction	IP allowlist/denylist
request-id	Generate unique request ID
real-ip	Extract real client IP from proxy headers
api-breaker	Circuit breaker for upstream failures
traffic-split	Traffic splitting for canary/A-B deployments

gRPC Support

Plugin	Description
grpc-transcode	Transcode HTTP/JSON to gRPC
grpc-web	gRPC-Web protocol support

Logging

Plugin	Description
http-logger	Send access logs via HTTP
tcp-logger	Send access logs via TCP
udp-logger	Send access logs via UDP

Observability

Plugin	Description
prometheus	Expose Prometheus metrics
opentelemetry	Distributed tracing via OTel

Extension

Plugin	Description
ext-plugin-pre-req	External plugin runner (pre-request)
ext-plugin-post-resp	External plugin runner (post-response)
serverless-pre-function	Inline Lua pre-request
serverless-post-function	Inline Lua post-response

Custom Plugins

tenant-rate-limit

Per-tenant sliding window rate limiter backed by Redis. Runs in the access phase at priority 1001 (after auth, before most other plugins).

Source: layers/api-gateway/plugins/tenant-rate-limit/

Schema (from schema.lua):

Parameter	Type	Default	Description
redis_host	string	"gw-redis"	Redis hostname
redis_port	integer	6379	Redis port
redis_timeout	integer	1000	Redis connection timeout (ms)
redis_pool_size	integer	100	Redis connection pool size
default_count	integer	100	Default request limit per window
time_window	integer	60	Time window in seconds
rejected_code	integer	429	HTTP status code for rate-limited requests
rejected_msg	string	"Rate limit exceeded for tenant"	Error message for rate-limited requests
tenant_header	string	"X-Tenant-ID"	Header to extract tenant_id from
allow_missing_tenant	boolean	false	If false, reject requests without tenant_id
missing_tenant_code	integer	401	HTTP status code when tenant_id is missing
show_limit_header	boolean	true	Include X-RateLimit-* headers in response

Redis key patterns used:

• rl:{tenant_id}:{route_id}:{window} -- sliding window sorted set
• rl:config:{tenant_id} -- per-tenant limit override

Response headers (when show_limit_header is true):

• X-RateLimit-Limit -- max requests per window
• X-RateLimit-Remaining -- remaining requests
• X-RateLimit-Reset -- Unix timestamp when window resets
• X-Tenant-ID -- tenant identifier

Behavior:

• Uses an atomic Redis Lua script for sliding window counting (single round-trip)
• Checks for per-tenant limit override in rl:config:{tenant_id} before applying default
• Fails open if Redis is unreachable (requests are allowed through)
• Caches the Redis script SHA for subsequent calls via EVALSHA

Route configuration example:

curl http://localhost:9180/apisix/admin/routes/1 \
  -H "X-API-KEY: $APISIX_ADMIN_KEY" \
  -X PUT -d '{
    "uri": "/api/*",
    "upstream": {"type": "roundrobin", "nodes": {"ai-gateway:4000": 1}},
    "plugins": {
      "tenant-rate-limit": {
        "default_count": 500,
        "time_window": 60,
        "redis_host": "gw-redis",
        "redis_port": 6379
      }
    }
  }'

clickhouse-logger

Batched request log writer to ClickHouse via the HTTP interface. Runs in the log phase at priority 399.

Source: layers/api-gateway/plugins/clickhouse-logger/init.lua

Schema:

Parameter	Type	Default	Description
clickhouse_url	string	"http://clickhouse:8123"	ClickHouse HTTP endpoint
database	string	"gateway"	Target database
table	string	"request_log"	Target table (the Buffer table)
timeout	integer	5000	HTTP request timeout (ms)
batch_max_size	integer	100	Flush buffer when this many entries accumulate
inactive_timeout	integer	5	(Schema default; periodic flush runs every 5 seconds)
buffer_duration	integer	60	(Schema default for buffer duration)
max_retry_count	integer	3	Max retries on write failure
retry_delay	integer	1	Delay between retries (seconds)
tenant_header	string	"X-Tenant-ID"	Header to extract tenant_id from

Behavior:

• Buffers log entries per Nginx worker process
• Flushes when buffer reaches batch_max_size or every 5 seconds (timer)
• Writes to ClickHouse using INSERT INTO ... FORMAT JSONEachRow
• Extracts tenant_id from header, rate-limit plugin context, or JWT payload
• Scrubs Authorization header -- never logs tokens or credentials
• Log entry fields: timestamp, tenant_id, route_id, method, path, status_code, latency_ms, request_size, response_size, client_ip, user_agent, trace_id

Route configuration example:

curl http://localhost:9180/apisix/admin/routes/1 \
  -H "X-API-KEY: $APISIX_ADMIN_KEY" \
  -X PATCH -d '{
    "plugins": {
      "clickhouse-logger": {
        "clickhouse_url": "http://clickhouse:8123",
        "database": "gateway",
        "table": "request_log",
        "batch_max_size": 200
      }
    }
  }'

Plugin Attributes

Plugin attributes configure global behavior for plugins (shared across all routes).

prometheus

plugin_attr:
  prometheus:
    export_uri: /apisix/prometheus/metrics
    export_addr:
      ip: 0.0.0.0
      port: 9091

Exposes Prometheus metrics on a dedicated port (9091) separate from the main proxy port. Scraped by Prometheus at apisix:9091.

opentelemetry

plugin_attr:
  opentelemetry:
    resource:
      service.name: apisix
    collector:
      address: otel-collector:4317
      request_timeout: 3
    trace_id_source: x-request-id
    batch_span_processor:
      max_queue_size: 1024
      batch_timeout: 2
      inactive_timeout: 1
      max_export_batch_size: 16

Setting	Value	Description
resource.service.name	apisix	Service name in traces
collector.address	otel-collector:4317	OTel Collector gRPC endpoint
collector.request_timeout	3	Timeout for sending spans (seconds)
trace_id_source	x-request-id	Use X-Request-ID header as trace ID
batch_span_processor.max_queue_size	1024	Max queued spans before drop
batch_span_processor.batch_timeout	2	Max wait before sending a batch (seconds)
batch_span_processor.inactive_timeout	1	Timeout for inactive batches (seconds)
batch_span_processor.max_export_batch_size	16	Max spans per export batch

http-logger

plugin_attr:
  http-logger:
    inactive_timeout: 2
    batch_max_size: 1000
    buffer_duration: 60

Setting	Value	Description
inactive_timeout	2	Flush after 2 seconds of inactivity
batch_max_size	1000	Max entries per batch
buffer_duration	60	Max buffer duration in seconds

Custom Plugin Directory Structure

Custom plugins are loaded from the extra_lua_path directory. Each plugin follows this structure:

layers/api-gateway/plugins/
  tenant-rate-limit/
    handler.lua      # Plugin lifecycle (access phase)
    schema.lua       # JSON schema for config validation
    init.lua         # Alternative entry point (same logic)
  clickhouse-logger/
    init.lua         # Plugin lifecycle (log phase) + schema

:::note Plugins must be listed in the plugins array in apisix.yaml to be loadable. Adding a plugin directory without listing it in the config will not enable it. :::