Index writer memory Continue to rise

Hi,
@nhat I still have a question,I use logstash double write data to es6.3 and es7.4 cluster.I found 6.3cluster also have a problematic index which is same as es7.4‘s problematic index.But 6.3 cluster is fine.Is there any difference about idnex writer in es6.3 and es7.4.
See the wrong document below:
image

Hi @ITzhangqiang, I believe the 6.3 cluster worked unexpectedly because of the regression bug. Previously, the indexing memory of each IndexWriter was incorrectly capped at 256MB. This cap was a regression and fixed in https://github.com/elastic/elasticsearch/pull/31105. With this fix, the problematic index in your 7.4 cluster could use up to 2GB instead of 256MB.