"Pit" of mongoreplay

This is how the story begins , Because I want to be a MONGODB The test library synchronizes the data of some production libraries , At the same time, I don't want to start from the basic data , Because of the large amount of data . In fact, it can be through mongoexport and mongoimport Come on , But the problem is, if you just want to capture data in a certain period of time , Or just want to reproduce the fault data in a certain period of time in the production library . How can we track these data and record , Replicate on the test library .

Actually from MONGODB 3.4 In the future, there will be a tool to help ,Mongoreplay, His main function is , Grab the packet through the network , Change the data intercepted through the network card into a specific data format , And then through   replay To replay data on another library , We can think of him as something like mysql Of binlog.  Of course, the principle may be different .

Where is the pit , Keep listening . When capturing data , To specify a network card , To specify the port number you want to capture

mongoreplay record -i ens32 -e "port 27027" -b 20 -p /mongodata/mongo.json

on the basis of BSON In the file you specify . Then you take this file , You can pour data into the specified MONGODB The server

Data source location , perform play Command can inject data into the test machine .

Where is the pit , During the test , Data is sometimes found without , And sometimes the data is irregular , Any command executed on the local machine cannot be found , All commands not executed on the local machine have related data fetching .

This involves , It's not about this command , It's the problem of capturing packets , In command -i The latter is the network card that must specify the data source , General machines have lo Local loop , and External physical network card . While monitoring , It is monitoring the external physical network card , The local commands are executed in lo In the circuit .

But here comes the problem , If you want a complete data pass mongoreplay It is difficult to command , Even if we produce two sets mongoreplay To grab local and Extranet data , However, we cannot identify the execution sequence in playback .

therefore mongoreplay This function , Only limited to system debugging and fault replay , For data synchronization , This function is really chicken ribs .