Bug Report

Previously, we only ensured compatibility issues when PD was upgraded first. If the scheduling server completed the upgrade first but PD has not been upgraded yet, there will be duplicate IDs.

1. Assume that the current PD base ID is 10.
2. The upgraded scheduling server receives the AskBatchSplit request. Assume that the request count is 1.
3. But the PD server has not been upgraded at this time, it is sent to the PD server for processing as a non-batch request. The base ID returned to the caller is 11, the region ID is 11, and the peer IDs are 12, 13, and 14 respectively.
4. After the PD server upgrade is completed, the scheduling server receives the AskBatchSplit request again.
5. Assuming that the request count is 1, it is processed as a batch request. And the base ID returned to the caller is 15-4=11, the region ID is 11, and the peer IDs are 12, 13, and 14.

So the bug is triggered that the region ID is the same but the range is different.