The labels are hierarchical, for example, . You can declare their hierarchies in the PD configuration file or pd-ctl:

• PD configuration file:

• pd-ctl:

  The number of machines must be no less than the .

Assume that the topology has three layers: zone > rack > host. You can set a label for each layer by command line parameter or configuration file, then TiKV reports its label to PD:

• TiKV command line parameter:

• TiKV configuration file:

  [server]
  

PD makes optimal scheduling according to the topological information. You only need to care about what kind of topology can achieve the desired effect.

Assume that you have 4 data zones, each zone has 2 racks, and each rack has 2 hosts. You can start 2 TiKV instances on each host as follows:

Start TiKV:

Configure PD:

$ pd-ctl

Now, PD schedules replicas of the same Region to different data zones.

• Even if one data zone goes down, the TiKV cluster is still highly available.
• If the data zone cannot recover within a period of time, PD removes the replica from this data zone.