TiFlash Issues

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

Original topic: tiflash问题

| username: jiangh

Currently, there is a TiFlash node in TiDB. After TiFlash is shut down, executing SQL involving tables stored in TiFlash results in: ERROR 1105 (HY000): close of nil channel. Querying information_schema.tiflash_replica also shows that the tables in TiFlash are all available. If this is used in production, the experience would be greatly affected if the TiFlash node goes down. Can’t it automatically switch to TiKV for queries?

| username: jiangh | Original post link

Sometimes it also reports: ERROR 9012 (HY000): TiFlash server timeout

| username: 啦啦啦啦啦 | Original post link

If a table has only a single TiFlash replica and the relevant node is unable to serve, queries in intelligent selection mode will keep retrying. You need to specify the Engine or manually use a Hint to read the TiKV replica. The purpose should be to prioritize TP business; otherwise, directing everything to TiKV might affect TP business. Therefore, it is recommended to have 2 TiFlash nodes in a production environment to maintain high availability.

| username: jiangh | Original post link

Does TiFlash high availability require additional high availability configuration, or is it sufficient to set the TIFLASH REPLICA count to 2? Will the two replicas be automatically allocated to two TiFlash nodes?

| username: 啦啦啦啦啦 | Original post link

No additional configuration is needed. Just set up 2 nodes and 2 replicas, and it will automatically distribute across all nodes.

| username: WalterWj | Original post link

It is possible to fall back, but it is not recommended. 系统变量 | PingCAP 文档中心

| username: system | Original post link

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