br备份失败,checksum的时候报错 Error: other error: Coprocessor task terminated due to exceeding the deadline

【 TiDB 使用环境】测试
【 TiDB 版本】6.5.0
【复现路径】br备份,1.2T数据量,两小时后失败
【遇到的问题:问题现象及影响】

使用br工具往共享存储中写入备份,2小时后失败
tiup br backup full --pd 192.168.40.93:2379 -s /share/mountpoint/s_tidb/tidb-test/87B21CECE46F4974943567BBC53253DE/

br日志
<-------------------------------------------------------------------------------> 99.71%Full Backup <------------------------------------------------------------------------------> 99.71%Full Backup <------------------------------------------------------------------------------|> 99.85%Full Backup <------------------------------------------------------------------------------/> 99.85%Full Backup <-------------------------------------------------------------------------------> 99.85%Full Backup <------------------------------------------------------------------------------> 99.85%Full Backup <------------------------------------------------------------------------------|> 99.85%Full Backup <------------------------------------------------------------------------------/> 99.85%Full Backup <------------------------------------------------------------------------------> 100.00%Checksum <…> 0.00%Checksum <…> 0.00%Checksum <…> 0.00%Checksum <…> 0.00%Checksum <–…> 1.46%Checksum ← .…> 1.46%Checksum <-|…> 1.46%Checksum <-/…> 1.46%Checksum <–…> 1.46%Checksum ← .…> 1.46%Checksum <-|…> 2.05%Checksum <-/…> 2.05%Checksum <–…> 2.05%Checksum ← .…> 2.05%Checksum <-|…> 2.05%Checksum .…> 2.05%Checksum <-|…> 2.05%[2024/06/11 20:19:10.909 +08:00] [INFO] [collector.go:73] [“Full Backup failed summary”] [total-ranges=21408] [ranges-succeed=21408] [ranges-failed=0] [backup-total-ranges=679]
Checksum <-/…> 2.05%Error: other error: Coprocessor task terminated due to exceeding the deadline

测试环境信息


br日志.txt (581.3 KB)

端口通么? 感觉在调用 tikv 的 API 做 checksum 计算 :thinking:TiKV | Checksum
然后超时了,可能是不通

是不是服务器配置太低了?

升级一下版本呢

你说的是升级数据库版本吗

业务端会有超时的报错 (Coprocessor task terminated due to exceeding the deadline)。通常这种情况都是执行计划不优化造成的,比如说缺少索引,导致需要全表扫描。