While the HTTP POST request succeeded and the received message was written into the journal, it may fail at a later stage in the processing (e. g. being dropped in a pipeline rule) or fail to index into Elasticsearch. All of that happens much after the client sent the HTTP request.
so for my case, should the Input already got those messages so there is no drops from input drops statistic, and where should i find out the related drops statistic ?
but it quite strange is that after i using a shell script to keep sending, messages got in ! and then i stop the shell, send message one by one manually, those messages also got in!
It seems there is a buffer on graylog to get filled and then works.