feature: add opentelemetry plugin. (apache#3891)

apisix/plugins/opentelemetry.lua

@@ -50,31 +50,35 @@ local pairs   = pairs
 local ipairs  = ipairs
 local unpack  = unpack
 
-local hostname
+local lrucache = core.lrucache.new({
+    type = 'plugin', count = 128, ttl = 24 * 60 * 60,
+})
+
 
 local attr_schema = {
     type = "object",
     properties = {
-        x_request_id_as_trace_id = {
-            type = "boolean",
-            description = "use x-request-id as new trace id",
-            default = false,
+        trace_id_source = {
+            type = "string",
+            enum = {"x-request-id", "random"},
+            description = "alternate use x-request-id as trace id",
+            default = "random",
         },
         resource = {
             type = "object",
             description = "additional resource",
-            additional_properties = {{type = "boolean"}, {type = "number"}, {type = "string"}},
+            additionalProperties = {{type = "boolean"}, {type = "number"}, {type = "string"}},
         },
         collector = {
             type = "object",
-            description = "otel collector",
+            description = "opentelemetry collector",
             properties = {
                 address = {type = "string", description = "host:port", default = "127.0.0.1:4317"},
                 request_timeout = {type = "integer", description = "second uint", default = 3},
                 request_headers = {
                     type = "object",
                     description = "http headers",
-                    additional_properties = {
+                    additionalProperties = {
                         one_of = {{type = "boolean"},{type = "number"}, {type = "string"}},
                    },
                 }
@@ -160,24 +164,17 @@ local schema = {
             },
             default = {name = "always_off", options = {fraction = 0, root = {name = "always_off"}}}
         },
-        tags = {
+        additional_attributes = {
             type = "array",
             items = {
-                type = "object",
-                properties = {
-                    position = {
-                        type = "string",
-                        enum = {"http", "arg", "cookie"}
-                    },
-                    name = {
-                        type = "string", minLength = 1
-                    }
-                }
+                type = "string",
+                minLength = 1,
             }
         }
     }
 }
 
+
 local _M = {
     version = 0.1,
     priority = -1200, -- last running plugin, but before serverless post func
@@ -186,10 +183,13 @@ local _M = {
     attr_schema = attr_schema,
 }
 
+
 function _M.check_schema(conf)
     return core.schema.check(schema, conf)
 end
 
+
+local hostname
 local sampler_factory
 local plugin_info
 
@@ -214,22 +214,16 @@ function _M.init()
         return
     end
 
-    if plugin_info.x_request_id_as_trace_id then
+    if plugin_info.trace_id_source == "x-request-id" then
         id_generator.new_ids = function()
             local trace_id = ngx_req.get_headers()["x-request-id"] or ngx_var.request_id
             return trace_id, id_generator.new_span_id()
         end
     end
 end
 
-local tracers = {}
-
-local function fetch_tracer(conf, ctx)
-    local t = tracers[ctx.route_id]
-    if t and t.v == ctx.conf_version then
-        return t.tracer
-    end
 
+local function create_tracer_obj(conf)
     -- create exporter
     local exporter = otlp_exporter_new(exporter_client_new(plugin_info.collector.address,
                                                             plugin_info.collector.request_timeout,
@@ -276,30 +270,33 @@ local function fetch_tracer(conf, ctx)
         sampler = sampler,
     })
     -- create tracer
-    local tracer = tp:tracer("opentelemetry-lua")
-    tracers[ctx.route_id] = {tracer = tracer, v = ctx.conf_version}
-
-    return tracer
+    return tp:tracer("opentelemetry-lua")
 end
 
+
 function _M.access(conf, api_ctx)
+    local tracer, err = core.lrucache.plugin_ctx(lrucache, api_ctx, nil, create_tracer_obj, conf)
+    if not tracer then
+        core.log.error("failed to fetch tracer object: ", err)
+        return
+    end
+
     -- extract trace context from the headers of downstream HTTP request
     local upstream_context = trace_context.extract(context, carrier_new())
     local attributes = {
         attr.string("service", api_ctx.service_name),
         attr.string("route", api_ctx.route_name),
     }
-    if conf.tags then
-        for _, tag in ipairs(conf.tags) do
-            local key = tag.position .. "_" .. tag.name
+    if conf.additional_attributes then
+        for _, key in ipairs(conf.additional_attributes) do
             local val = api_ctx.var[key]
             if val then
                 core.table.insert(attributes, attr.string(key, val))
             end
         end
     end
 
-    local ctx, _ = fetch_tracer(conf, api_ctx):start(upstream_context, api_ctx.var.request_uri, {
+    local ctx, _ = tracer:start(upstream_context, api_ctx.var.request_uri, {
         kind = span_kind.client,
         attributes = attributes,
     })
@@ -309,6 +306,7 @@ function _M.access(conf, api_ctx)
     trace_context.inject(ctx, carrier_new())
 end
 
+
 function _M.body_filter(conf, ctx)
     if ngx.arg[2] then
         local upstream_status = core.response.get_upstream_status(ctx)
@@ -323,14 +321,5 @@ function _M.body_filter(conf, ctx)
     end
 end
 
-function _M.destroy()
-    if process.type() ~= "worker" then
-        return
-    end
-
-    for _, t in pairs(tracers) do
-        t.tracer.provider:shutdown()
-    end
-end
 
 return _M

conf/config-default.yaml

@@ -379,6 +379,7 @@ plugins:                          # plugin list (sorted by priority)
   # <- recommend to use priority (0, 100) for your custom plugins
   - example-plugin                 # priority: 0
   #- skywalking                    # priority: -1100
+  #- opentelemetry                 # priority: -1200
   - aws-lambda                     # priority: -1899
   - azure-functions                # priority: -1900
   - openwhisk                      # priority: -1901

docs/en/latest/plugins/opentelemetry.md

@@ -47,9 +47,8 @@ Just support reporting in `HTTP` with `Content-Type=application/x-protobuf`, the
 | sampler.options.root.name | string | optional | always_off | ["always_on", "always_off", "trace_id_ratio"] | sampling strategy
 | sampler.options.root.options | object | optional | {fraction = 0} | | sampling strategy parameters
 | sampler.options.root.options.fraction | number | optional | 0 | [0, 1] | trace_id_ratio fraction
-| tags | array[object] | optional | | | append to trace span attributes
-| tags.position | string | required | | ["http", "arg", "cookie"] | where variable in
-| tags.name | string | required | | | variable name
+| additional_attributes | array[string] | optional | | | append to trace span attributes
+| additional_attributes[0] | string | required | | | key of ctx.var
 
 ## How To Enable
 
@@ -95,18 +94,18 @@ We can set the collecting by specifying the configuration in `conf/config.yaml`.
 
 | Name         | Type   | Default  | Description                                                          |
 | ------------ | ------ | -------- | ----------------------------------------------------- |
-| x_request_id_as_trace_id | boolean | false | use current request id as new TraceID, you should make sure the request id is match regex pattern: `[0-9a-f]{32}`|
+| trace_id_source | enum | random | alternate use x-request-id as trace id, valid value is `random` or `x-request-id`, if use `x-request-id`, please make sure it match regex pattern `[0-9a-f]{32}` |
 | resource | object |   | additional [resource](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/resource/sdk.md) append to trace |
 | collector | object | {address = "127.0.0.1:4317", request_timeout = 3} | otlp collector |
 | collector.address | string | 127.0.0.1:4317 | collector address |
-| collector.request_timeout | integer | 3 | report request timeout |
+| collector.request_timeout | integer | 3 | report request timeout(second) |
 | collector.request_headers | object |  | report request http headers |
 | batch_span_processor | object |  | trace span processor |
 | batch_span_processor.drop_on_queue_full | boolean | true | drop span when queue is full, otherwise force process batches |
 | batch_span_processor.max_queue_size | integer | 2048 | maximum queue size to buffer spans for delayed processing |
 | batch_span_processor.batch_timeout | number | 5 | maximum duration(second) for constructing a batch |
 | batch_span_processor.max_export_batch_size | integer | 256 | maximum number of spans to process in a single batch |
-| batch_span_processor.inactive_timeout | number | 2 | timer interval for processing batches |
+| batch_span_processor.inactive_timeout | number | 2 | timer interval(second) for processing batches |
 
 Here is an example:
 

docs/zh/latest/plugins/opentelemetry.md

@@ -95,18 +95,18 @@ curl http://127.0.0.1:9080/apisix/admin/routes/1  -H 'X-API-KEY: edd1c9f034335f1
 
 | 名称         | 类型   | 默认值   | 描述                                                  |
 | ------------ | ------ | -------- | ----------------------------------------------------- |
-| x_request_id_as_trace_id | boolean | false | 使用当前请求 ID 作为新的 TraceID，必须确保当前请求 ID 是符合 TraceID 规范的：`[0-9a-f]{32}` |
+| trace_id_source | enum | random | 合法的取值：`random` 或 `x-request-id`，允许使用当前请求 ID 代替随机 ID 作为新的 TraceID，必须确保当前请求 ID 是符合 TraceID 规范的：`[0-9a-f]{32}` |
 | resource | object |   | 追加到 trace 的额外 [resource](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/resource/sdk.md) |
 | collector | object | {address = "127.0.0.1:4317", request_timeout = 3} | 数据采集服务 |
 | collector.address | string | 127.0.0.1:4317 | 数据采集服务地址 |
-| collector.request_timeout | integer | 3 | 数据采集服务上报请求超时时长 |
+| collector.request_timeout | integer | 3 | 数据采集服务上报请求超时时长，单位秒 |
 | collector.request_headers | object |  | 数据采集服务上报请求附加的 HTTP 请求头 |
 | batch_span_processor | object |  | trace span 处理器参数配置 |
 | batch_span_processor.drop_on_queue_full | boolean | true | 当处理器缓存队列慢试，丢弃新到来的 span |
 | batch_span_processor.max_queue_size | integer | 2048 | 处理器缓存队列容量最大值 |
-| batch_span_processor.batch_timeout | number | 5 | 单位秒，构造一批 span 超时时长 |
+| batch_span_processor.batch_timeout | number | 5 | 构造一批 span 超时时长，单位秒 |
 | batch_span_processor.max_export_batch_size | integer | 256 | 一批 span 的数量，每次上报的 span 数量 |
-| batch_span_processor.inactive_timeout | number | 2 | 每隔多长时间检查是否有一批 span 可以上报 |
+| batch_span_processor.inactive_timeout | number | 2 | 每隔多长时间检查是否有一批 span 可以上报，单位秒 |
 
 配置示例:
 

t/plugin/opentelemetry.t

@@ -128,8 +128,6 @@ GET /opentracing
 --- response_body
 opentracing
 --- wait: 1
---- no_error_log
-[error]
 --- grep_error_log eval
 qr/opentelemetry export span/
 --- grep_error_log_out
@@ -195,8 +193,6 @@ opentracing
 --- grep_error_log eval
 qr/opentelemetry export span/
 --- grep_error_log_out
---- no_error_log
-[error]
 
 
 
@@ -264,8 +260,6 @@ opentracing
 --- grep_error_log eval
 qr/opentelemetry export span/
 --- grep_error_log_out
---- no_error_log
-[error]
 
 
 
@@ -337,8 +331,6 @@ GET /opentracing
 --- response_body
 opentracing
 --- wait: 1
---- no_error_log
-[error]
 --- grep_error_log eval
 qr/opentelemetry export span/
 --- grep_error_log_out
@@ -410,8 +402,6 @@ opentracing
 --- grep_error_log eval
 qr/opentelemetry export span/
 --- grep_error_log_out
---- no_error_log
-[error]
 
 
 
@@ -487,8 +477,6 @@ GET /opentracing
 --- response_body
 opentracing
 --- wait: 1
---- no_error_log
-[error]
 --- grep_error_log eval
 qr/opentelemetry export span/
 --- grep_error_log_out
@@ -570,8 +558,6 @@ opentracing
 --- grep_error_log eval
 qr/opentelemetry export span/
 --- grep_error_log_out
---- no_error_log
-[error]
 
 
 
@@ -583,8 +569,6 @@ traceparent: 00-00000000000000000000000000000001-0000000000000001-01
 --- response_body
 opentracing
 --- wait: 1
---- no_error_log
-[error]
 --- grep_error_log eval
 qr/opentelemetry export span/
 --- grep_error_log_out
@@ -670,8 +654,6 @@ GET /opentracing
 --- response_body
 opentracing
 --- wait: 1
---- no_error_log
-[error]
 --- grep_error_log eval
 qr/opentelemetry export span/
 --- grep_error_log_out
@@ -689,5 +671,3 @@ opentracing
 --- grep_error_log eval
 qr/opentelemetry export span/
 --- grep_error_log_out
---- no_error_log
-[error]