|
299 | 299 | %% See https://www.rabbitmq.com/ha.html#queue-master-location for further details
|
300 | 300 | %% {queue_master_locator, <<"client-local">>},
|
301 | 301 |
|
| 302 | + %% Since RabbitMQ 3.6.0, masters perform synchronisation in batches. |
| 303 | + %% Batch can be configured via the ha-sync-batch-size queue argument. |
| 304 | + %% Earlier versions will will synchronise 1 message at a time by default. |
| 305 | + %% By synchronising messages in batches, the synchronisation process can be sped up considerably. |
| 306 | + %% To choose the right value for ha-sync-batch-size you need to consider: |
| 307 | + %% * average message size |
| 308 | + %% * network throughput between RabbitMQ nodes |
| 309 | + %% * net_ticktime value |
| 310 | + %% For example, if you set ha-sync-batch-size to 50000 messages, |
| 311 | + %% and each message in the queue is 1KB, |
| 312 | + %% then each synchronisation message between nodes will be ~49MB. |
| 313 | + %% You need to make sure that your network between queue mirrors can accomodate this kind of traffic. |
| 314 | + %% If the network takes longer than net_ticktime to send one batch of messages, |
| 315 | + %% then nodes in the cluster could think they are in the presence of a network partition. |
| 316 | + %% {mirroring_sync_batch_size, 4096}, |
| 317 | + |
302 | 318 | %% To announce custom properties to clients on connection:
|
303 | 319 | %%
|
304 | 320 | %% {server_properties, []},
|
|
0 commit comments