What do pd cache and cluster cache specifically refer to?

Translated
1

Note:
This topic has been translated from a Chinese forum by GPT and might contain errors.

Original topic: pd的cache，cluster cacne这些，具体是指代那些呢

| username: Lystorm

I have some questions about the cache involved in the PD module and would like to ask for some clarification from everyone.

  1. Is the cache mentioned here the cache within the PD cluster?

    image
    image976×654 79.6 KB

  2. Is this cluster cache the cache in the PD Client within the TiDB Service?

    image
    image1368×777 132 KB

  3. Is the cache in the PD Client within the TiDB Service actively updated by TiDB, or is it updated by PD?

2
| username: forever | Original post link
  1. It is the cache of the PD cluster.
  2. Here, TiKV reports updates to the cache of the PD cluster.
  3. In the TiDB Service, the cache in the PD Client will back off and retrieve from PD again to update the PD client cache when the region information is incorrect.
3
| username: Lystorm | Original post link

Got it.
I would like to ask again, does each PD node run an entire etcd cluster or a single etcd node?

4
| username: OnTheRoad | Original post link

Each PD node runs one etcd instance.

5
| username: cs58_dba | Original post link

The cluster is definitely deployed in a distributed manner.

6
| username: Lystorm | Original post link

This means that each PD contains an etcd instance, and these instances form an etcd cluster. PD works as a single point, so according to this etcd deployment logic, the data recorded in each PD’s etcd is incomplete?

7
| username: forever | Original post link

This is a cluster, and theoretically, all etcd nodes have the same data.

8
| username: Lystorm | Original post link

Does a single etcd instance contain all the data?

9
| username: xiaohetao | Original post link

Well, one of the instances is the leader.

10
| username: cs58_dba | Original post link

Three data replicas are required for high availability.

11
| username: forever | Original post link

etcd maintains communication and data consistency among the nodes in the cluster through the Raft algorithm. The nodes are in a peer-to-peer relationship, keeping the same data. Even if the leader node fails, a new leader will be quickly elected to ensure the normal operation of the system.

12
| username: system | Original post link

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.