Closed wyukawa closed 5 years ago
This problem doesn't happen in fluent-plugin-elasticsearch 3.5.2, elasticsearch ruby 7.1.0, Elasticsearch 7.1.1
Please increase 'chunk_limit_size'
@Z0T0 You shouldn't write your issue report in non-relevant issue. Please open a new issue instead.
Problem
We send logs our elasticsearch with fluent-plugin-elasticsearch but deadlock error occurred when we restarted our elasticsearch.
I'm not sure this problem is related to fluent-plugin-elasticsearch or elasticsearch-ruby.
This error also occurred in fluent-plugin-elasticsearch 2.11.10.
At that time, I thought this was related to elasticsearch-ruby. So I filed https://github.com/elastic/elasticsearch-ruby/issues/656
At least, we don't have the problem when we use fluent-plugin-elasticsearch 2.11.10 and Elasticsearch 6.7.1
Steps to replicate
data flow is here kafka -> kafka-fluentd-consumer -> fluend v1 -> fluent-plugin-elasticsearch -> elasticsearch
fluentd conf
Expected Behavior or What you need to ask
No deadlock
If deadlock occurs, fluent-plugin-elasticsearch retries but fails. So buffer queue length increases.
Using Fluentd and ES plugin versions
Gemfile
gem "fluentd", "1.4.2" gem "oj", "3.7.12"
gem "fluent-plugin-ping-message", "1.0.0" gem "fluent-plugin-record-reformer", "0.9.1" gem "fluent-plugin-suppress", "1.1.0" gem 'fluent-plugin-elasticsearch', :git => 'https://github.com/uken/fluent-plugin-elasticsearch.git', :branch => " configurable-split_request_p" gem "typhoeus", "1.3.1" gem "fluent-plugin-prometheus", "1.3.0"