This way, all index have a regular size (which is a good practice I think) but I can't be sure that logs will be conserved one year because an index could take one hour, or three days or whatever to reach the size of 30 GB right ? Some could be conserved some month while other could be conserved more than one year
My only option would be to add max age : 1 to my hot phase, but index size will not be regular (i.e more logs on week and less in weekends)
So, what is your best strategy to have a good compromise between retention time and index size ?
I would recommend setting a max age for the hot phase of perhaps a week as this seems reasonable given your retention period. If this results in some indices that are smaller than the size target that should not be a problem.
Then adjust you retention settings so you always keep at least a years worth of data. If the oldest index covers a full week this means you at times will hold a year plus a week which in my experience generally is fine.
That does not matter. Moving shards between zones is one aspect of the index lifecycle, but there are other steps that apply even if you only have a single zone.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.