Thought i would propose a question to the "guru's" and see what the general thought would be......
I have a stream of data coming from some proprietary software of "transactions" which i want to put into ES. now this stream is through a java plugin which i write and control. In theory i can do almost anything with it. The two possible options i have come up with
1 take the data and use log4J and write to a log file. then from there i use logstash with FileBeat (filters etc) to get that data loaded into transactions
Pros: if logstash stops/ starts i dont lose any logs as it starts from where it last processed
Cons: extra puzzle pieces to break and have to get working.
more disk IO and resources being used to write/ read the log file
#2 in my plugin i use the elasticsearch rest api and or bulk api to push straight into ES
Pros : less to break and easier on resources (i think)
Cons: if ES goes down i will lose data in the stream but thats not "critical" to me as data is being used for a monitoring perspective so i don't need 100% of the data. Loss is acceptable.
On the volume side of things its quite low for ES at around 1 million records a day......
Interested to see what anyone thinks of my little problem
-Julian