dingshun3016

Results 17 comments of dingshun3016

> I don't see the root cause of this bug. @dingshun3016 Could you help add test case to simulate this or add more description I post one case log in...

> Could you check the value of the metric `total_dropped_event_num` in every shuffle server? I encountered "Blocks read inconsistent" exception in #1620 when there are events being dropped. > >...

> > > Could you check the value of the metric `total_dropped_event_num` in every shuffle server? I encountered "Blocks read inconsistent" exception in #1620 when there are events being dropped....

> Thanks for adding detailed logs, I have understood your point, the root cause is to partition shuffle reader handler is cached. After the memory data is flushed into local...

This function is very useful, but I'm a little worried. when all apps in the cluster are enabled, if server a1 is inactive, it will be immediately assigned to another...

When I set MAX_DIRECT_MEMORY_SIZE=30G, jvm_buffer_pool_used_bytes{pool="direct"} can reach 30G ![截屏2024-01-15 18 37 38](https://github.com/apache/incubator-uniffle/assets/21172339/8c69da99-4366-4a8d-a31f-6e3c27c82ead)

Do you have any ideas about this issue? @zuston

> @dingshun3016 Could you create an issue for this? Okay, I've created

> Why we need jetty package in client side? @dingshun3016 You can take a look at the stack information above, mainly using the MultiException class in jetty to determine whether...

I moved MultiException from the jetty package to uniffle and renamed it to RssMultiException