Node catalog is not ready, please wait for a while

Viewed 115

我们集群现在有一个hive catalog,五个mysql catalog,还有一个doris内部的catalog,mysql和hive都有设置刷新间隔

问题:
1、这个问题是在执行什么操作的时候会发生?
2、应该如何避免这个问题?

报错信息如下:
ERROR 1105 (HY000) at line 2: Exception, msg: java.lang.Exception: Node catalog is not ready, please wait for a while. To master progress: .
Frontends:
name: fe_efde7dcd_525d_423c_a6df_cd2dddc88259, role: FOLLOWER, ali-doris-fe-prod-02:9010, is alive: true
name: fe_b01c8ade_b7ef_4037_af75_66945277ec12, role: FOLLOWER, ali-doris-fe-prod-03:9010, is alive: true
name: fe_cfb95a22_14b4_4df1_9fc1_6b4e930dea57, role: FOLLOWER, ali-doris-fe-prod-01:9010, is alive: true
Electable nodes:
ali-doris-fe-prod-02/...:9010
ali-doris-fe-prod-03/...:9010
ali-doris-fe-prod-01/...:9010

2 Answers

通常是由于master写入比较频繁,导致follower一定时间内replayer消费不了
,follower的replayer如果发现当前时间和上一条的synchronizedTimeMs差太多,设置自己not ready状态导致;可以考虑将follower的meta_delay_toleration_second参数调整大。

如果外部数据源多没有频繁的元数据变动,不建议设置太频繁的 catalog 定时刷新频次