Replies: 5 comments 4 replies
-
@maben996 2.3.0 及以上全量阶段是可以多并行度同步的, |
Beta Was this translation helpful? Give feedback.
-
@gtk96 实测了一下,只有当'scan.incremental.snapshot.enabled' = 'true' 和 'scan.startup.mode' = 'initial' 都配置的时候,source回使用默认并行度,只开'scan.startup.mode' = 'initial' source并行度还是1。 |
Beta Was this translation helpful? Give feedback.
-
增加资源会不会快一点呢? |
Beta Was this translation helpful? Give feedback.
-
理论上讲,source 端没有性能瓶颈的,瓶颈主要是 sink 端,所以你在 source 资源,包括拉取数据量都可以放大一些 |
Beta Was this translation helpful? Give feedback.
-
问题解决了么? |
Beta Was this translation helpful? Give feedback.
-
UI上看到QPS在2万左右,并且source的并行度一直是1。请问如何提高source的并行度?以及并行度的提升是否有助于提升消费速率?
sql 的 source table 配置如下,这里在CDC文档中没有找到可以配置source并行度的地方。提高了拉数据的batch size。发现对source的QPS没有提升效果。
flink.conf文件中的默认并行度配置如下
业务背景:有一批百亿级别的数据需要同步从mongo同步到doris,存量+增量都要同步,所以想使用CDC同步,但是发现速率太慢,目前QPS 2w,预计300亿要同步20天左右才能跑完。
Beta Was this translation helpful? Give feedback.
All reactions