Defining a custom ES index based on an existing field

I have a logstash pipeline designed to break container logs coming from coreos into separate indexes, but it's not quite working:

My pipeline looks like this:

input {
  beats {
    port => 5044
  }
}

filter {
  grok {
    match => { "source" => "%{GREEDYDATA}/%{CONTAINERAPP:prefix}*" }
    patterns_dir => ["/patterns"]
  }
......
  translate {
    add_tag => [ "monitoring" ]
    field => "prefix"
    regex => true
    exact => true
    dictionary_path => "/dict/monitoring.yml"
    destination => "container"
  }
.......
[Several more blocks like this adding tags to different groups of "prefix" fields]
.......
}
output {
  elasticsearch {
    hosts => ["elasticsearch:9200"]
    manage_template => false
    index => "logtest-%{[@metadata][container]}-%{+YYYY.MM.dd}"
    document_type => "%{[@metadata][type]}"
    }
}

I'm trying to use the syntax suggested here to get logstash to write the events to an index based on the name of the container. However, what I'm ending up with are indexes with the literal string, rather than the contents of the field "container", which is what I'm after.

$ curl -v elasticsearch:9200/_cat/indices?v |grep logtest
.....
green  open   logtest-%{[@metadata][container]}-2017.05.02 unVVBghgRra974trYtDPJA   5   1       9168            0     18.9mb          9.5mb

This doesn't happen with certain other fields, for instance, up until now I have had the "type" field in the pipeline config:

.......
    index => "logtest-%{[@metadata][type]}-%{+YYYY.MM.dd}"
.......

and have ended up with indices like:

.......
green  open   logtest-log-2017.04.06                       EzNg2CMKSx-mMMpbDVF-9g   5   1  116131362            0    202.1gb          101gb
.......

Thanks in advance.

I'm not sure about this but @metadata isn't for fields like _type, etc.?
Did you try simply %{container}?

2 Likes

You are storing the data in the field container, not @metadata.container. Change the index name specification to"logtest-%{container}-%{+YYYY.MM.dd}". Having said that, be careful with this approach as you could easily end up with a lot of very small indices, which is very, very inefficient.

1 Like

Thanks for the very quick replies!

Both were along the right lines. It's now working.. :smiley::thumbsup:

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.