Logstash pipeline output | duplicate messages ending up indexes

[deeps]

In the /etc/logstash/conf.d/ directory I have configured 2 pipeline files to read messages from kafka topic/s and send it to data nodes in the cluster. (1.conf and 2.conf)

[root@ingest1 conf.d]# egrep -w "topics|index" * | uniq
1.conf:        topics => ["events"]
1.conf:    index => "events-%{+YYYY.MM.dd}"
2.conf:        topics => ["input"]
2.conf:    index => "input-%{+YYYY.MM.dd}"

But if I produce a message to "events" topic, the message is ending up in 2 indexes. Same with other topic also.
Am I missing anything?

[Christian_Dahlqvist]

All files in the config directory are concatenation into a smaller neglected pipeline, meaning that data from all inputs go through all filters and are sent to all outputs unless you use conditionals to control the glow.

[deeps]

@Christian_Dahlqvist thanks for the quick response.
could you also guide me how to solve this?
ex: i want 1.conf to read "events" topic, apply pipeline filter and just output to its index.

TIA!

[Badger]

If the two configurations are completely separate from input to output I would strongly suggest using multiple pipelines. If there is overlap, or you are stuck on an old version then you can use something like

add_field => { inputTopic => "events" }

(with two different value for inputTopic) on the inputs to distinguish them, then use

output {
    if [inputTopic] == "events" {
        elasticsearch {
             ...
        }
    }
}

to send them to different end-points.

Even better, since you are using a kafka input, you can have the input decorate the metadata with the topic name and then make the output configuration conditional upon that.

[deeps]

@Badger thanks for this info!
I ended up sending all the topics data to one index per the request from Dev's for easy searching.
Notes about the multiple pipelines was also easy to understand and implement.