Batch Processor

The batch processor can be used to aggregate entries(logs/any data) and process them in a batch. When the batch_max_size is set to 1 the processor will execute each entry immediately. Setting the batch max size more than 1 will start aggregating the entries until it reaches the max size or the timeout expires.

Configurations

The only mandatory parameter to create a batch processor is a function. The function will be executed when the batch reaches the max size or when the buffer duration exceeds.

NameTypeRequirementDefaultValidDescription
namestringoptionallogger’s name[“http logger”,…]A unique identifier used to identify the batch processor, which defaults to the name of the logger plug-in that calls the batch processor, such as plug-in “http logger” ‘s name is “http logger.
batch_max_sizeintegeroptional1000[1,…]Sets the maximum number of logs sent in each batch. When the number of logs reaches the set maximum, all logs will be automatically pushed to the HTTP/HTTPS service.
inactive_timeoutintegeroptional5[1,…]The maximum time to refresh the buffer (in seconds). When the maximum refresh time is reached, all logs will be automatically pushed to the HTTP/HTTPS service regardless of whether the number of logs in the buffer reaches the maximum number set.
buffer_durationintegeroptional60[1,…]Maximum age in seconds of the oldest entry in a batch before the batch must be processed.
max_retry_countintegeroptional0[0,…]Maximum number of retries before removing the entry from the processing pipeline when an error occurs.
retry_delayintegeroptional1[0,…]Number of seconds the process execution should be delayed if the execution fails.

The following code shows an example of how to use batch processor in your plugin:

  1. local bp_manager_mod = require("apisix.utils.batch-processor-manager")
  2. ...
  4. local plugin_name = "xxx-logger"
  5. local batch_processor_manager = bp_manager_mod.new(plugin_name)
  6. local schema = {...}
  7. local _M = {
  8.     ...
  9.     name = plugin_name,
  10.     schema = batch_processor_manager:wrap_schema(schema),
  11. }
  13. ...
  16. function _M.log(conf, ctx)
  17.     local entry = {...} -- data to log
  19.     if batch_processor_manager:add_entry(conf, entry) then
  20.         return
  21.     end
  22.     -- create a new processor if not found
  24.     -- entries is an array table of entry, which can be processed in batch
  25.     local func = function(entries)
  26.         -- serialize to json array core.json.encode(entries)
  27.         -- process/send data
  28.         return true
  29.         -- return false, err_msg, first_fail if failed
  30.         -- first_fail(optional) indicates first_fail-1 entries have been successfully processed
  31.         -- and during processing of entries[first_fail], the error occurred. So the batch processor
  32.         -- only retries for the entries having index >= first_fail as per the retry policy.
  33.     end
  34.     batch_processor_manager:add_entry_to_new_processor(conf, entry, ctx, func)
  35. end

The batch processor’s configuration will be set inside the plugin’s configuration. For example:

Batch Processor - 图1note

You can fetch the admin_key from config.yaml and save to an environment variable with the following command:

  1. admin_key=$(yq '.deployment.admin.admin_key[0].key' conf/config.yaml | sed 's/"//g')
  1. curl http://127.0.0.1:9180/apisix/admin/routes/1 -H "X-API-KEY: $admin_key" -X PUT -d '
  2. {
  3.       "plugins": {
  4.             "http-logger": {
  5.                 "uri": "http://mockbin.org/bin/:ID",
  6.                 "batch_max_size": 10,
  7.                 "max_retry_count": 1
  8.             }
  9.        },
  10.       "upstream": {
  11.            "type": "roundrobin",
  12.            "nodes": {
  13.                "127.0.0.1:1980": 1
  14.            }
  15.       },
  16.       "uri": "/hello"
  17. }'

If your plugin only uses one global batch processor, you can also use the processor directly:

  1. local entry = {...} -- data to log
  2. if log_buffer then
  3.     log_buffer:push(entry)
  4.     return
  5. end
  7. local config_bat = {
  8.     name = config.name,
  9.     retry_delay = config.retry_delay,
  10.     ...
  11. }
  13. local err
  14. -- entries is an array table of entry, which can be processed in batch
  15. local func = function(entries)
  16.     ...
  17.     return true
  18.     -- return false, err_msg, first_fail if failed
  19. end
  20. log_buffer, err = batch_processor:new(func, config_bat)
  22. if not log_buffer then
  23.     core.log.warn("error when creating the batch processor: ", err)
  24.     return
  25. end
  27. log_buffer:push(entry)

Note: Please make sure the batch max size (entry count) is within the limits of the function execution. The timer to flush the batch runs based on the inactive_timeout configuration. Thus, for optimal usage, keep the inactive_timeout smaller than the buffer_duration.