Note:
This topic has been translated from a Chinese forum by GPT and might contain errors.Original topic: Tidb如何实现类似只读取mysql从库的操作
|
username: Hacker_5dlWhpPx
[TiDB Usage Environment] Production Environment / Testing / PoC
[TiDB Version] 5.0 and above
[Encountered Problem] Deadlocks occur during query and insert modifications. I want to achieve read-write separation, for example, in MySQL, you can read data only from the replica. Can using follower read achieve a similar effect?
You can refer to this document:
Deadlock is a resource competition in business scenarios, and this requires considering a better solution. Relying on the database alone still cannot avoid it.
Thank you for your reply. According to the documentation, stale read is reading historical data. But I just want to simulate the operation similar to reading from a MySQL replica.
Or I won’t mention deadlocks. I just purely want to read replica data. Can it achieve a similar effect to reading from a MySQL replica?
Yes, to fully simulate, you can set up TiCDC, then connect it to a MySQL or another complete set of TiDB, and it will work. 
The cost is too high this way. Considering TiDB has replicas, I hope to only read data from the replicas.
Is there a way to specify that SQL only reads the follow replica? I only know the implementation principle of follow read.
You’re right, I posted this to ask if there is such a method.
This is a product of a synchronous process, it won’t work, right?
It should be necessary to find a way to solve the cause of the deadlock. If it’s read-only, there might be issues with accuracy.
Try synchronizing data to MySQL with TiCDC, and use Atlas or Cetus for read-write separation.
The deadlock issue might be a problem with the program, and implementing read-write separation probably won’t solve it.
|
username: 
|
username: 
|
username: 
|
username: 
|
username: