In order to follow up the processing flow of the balance&split of the chunk block separately, this is not covered here. But do not misunderstand that there is no association between the two.
read&write MongoDB give me the feeling is that a bowel access to the ground, all the processing is seemingly a bit of a single thread meaning. The schematic diagram is as follows:
processing Process: After processing Ⅰ, cluster startup, config server acquires the metadata of the entire cluster, preparing for subsequent addressing. #数据插入, UPDATE, delete the occasion, metadata will update, chunk data update before, will update the latest data to config server.
Request Processing ①, client send Read&write request ②, send request via driver to Router server ③, router Server resolves user's request ④, according to Config The metadata stored in the server, addressing shard corresponding chunk blocks, and the location of the target data ⑤, sending Shard to all associations Read&write request ⑥, individual shard feedback read&write result information ⑦, Query results are distributed across multiple Shard, router server merges query data
⑧, through driver feedback results information
➈, the client displays the latest results