Thanks ashutosh, Could I know where the logic implications exist, it is graph storage layer or kafka consumer layer? If it is kafka consumer, I my case, I can simply walk around it.
Thanks Ashutosh Mestry <[email protected]> 于2019年1月3日周四 上午2:23写道： > Hi > > The way things are right now, it is not possible to simply increase the > number of threads. There are logic implications that need to be considered > before multi-threaded approach can become feasible. > > We are in the process of adding number of improvements for improving rate > of ingestion. Please see this recent JIRA and see if this helps: > - ATLAS-3006 (https://issues.apache.org/jira/browse/ATLAS-3006). > > Best regards, > > ~ ashutosh > > > On 1/1/19, 11:32 PM, "严细浪" <[email protected]> wrote: > > We have a big cluster, after integrate atlas with our hive servers, > lots of > notification messages in Kafka are not processed because the event > process > is too slow. > > We increase atlas.notification.hook.numthreads but nothing changed. > Then we > read the code, noticed that the kafka consumer is hard code to 1 > thread. > My question is: > 1. why it is hard code as 1, is there limitation in atlas that there > could > be only one thread can update data > 2. can we increase thread count. > > >
