The purpose of the PD’s isolation-level configuration is to ensure that the replicas of a region are not all distributed within the range specified by the isolation-level. For example, if the isolation-level is set to zone, then the three replicas of a region cannot all be distributed within the same zone. Similarly, if the isolation-level is set to host, then the three replicas of a region cannot all be distributed within the same host. Is this understanding correct?
Personally, I think so too, and the related tutorials explain it this way.
Refer to the TiDB Database Core Principles and Architecture [TiDB v6] course Lesson 04 Placement Driver classroom quiz.
If the isolation-level is set to zone, then the three replicas of the region will be distributed in different zones.
If the isolation-level is set to host, then the three replicas of the region will be distributed in different hosts.
Even if adhering to this isolation restriction makes it impossible to meet the multi-replica requirement of max-replicas, PD will not perform the corresponding scheduling.
The isolation-level configuration parameter of PD is only used to specify the transaction isolation level of TiDB and has no direct relation to the distribution of Region replicas.
|
username: Raymond
|
username: 
|
username: 
|
username: 
|
username: 
|
username: 
|
username: