Making Logstash configurations dynamic

Elastic Stack Logstash
1

Is there a way to make Logstash configuration file dynamic by reading attribute names and types from a different text file (or any other input mechanism).

To explain, if my logstash configuration file looks something like this:

if[@metadata][f] == "SCALE" {
    csv {
      separator => ","
      columns => ["TIMESTAMP",	"Machine_ID", "FLAG",	"Batch_ID",	"Scale",	"Op_Name",	"Set_Point", "Avg_Wght", "PPM",	"T1PPM"]
    }
    mutate { add_field => { "[@metadata][target_index]" => "dev-mw-1" } }
  }

The column names in this configuration are hard coded and each of the field is converted into its respective type using mutate.

Can I take these column names and their data type from some other input source (e.g. a text file), save them into a variable and use that variable? So that whenever Logstash is run, it takes fresh attribute from input and changes the configurations accordingly?

2

The logstash configuration can reference environment variables. However, it does not work everywhere. Suppose we have

COL1=foo
COL2=bar
COL1TYPE=integer
COL2TYPE=integer

Then the configuration

input { generator { count => 1 lines => [ '123,456' ] } }
filter {
    csv { columns => [ "${COL1}", "${COL2}" ] }
    mutate { convert => { "${COL1}" => "${COL1TYPE}" } }
    mutate { convert => { "bar" => "${COL2TYPE}" } }
}
output  { stdout { codec => rubydebug { metadata => false } } }

will result in

       "bar" => 456,
       "foo" => "123",

Note that bar is an integer, but foo is not. The left side of the convert is not substituted. Also a string that looks like an array will be interpreted as a string, not an array. So you cannot do

COLUMNS=foo, bar

and then use it like this

csv { columns => [ ${COLUMNS} ] }