Thinking about running an operation procedure that should do flushed sync on daily 'rotated' indices and wondering...
Would this be a feasible and good idea and what to avoid?
If we after midnight do synced flush per index of 'yesterday' and avoid any future [re]flush operations on such indices, would this maintain sync markers per daily indices' shards for future faster resyncs/restarts?
Or should we just do a synced flush on all indices at demand eg. before performing a rolling restart etc.?
There is no need. In recent versions synced flush is deprecated and does nothing more than a regular flush, which Elasticsearch will do automatically anyway.
"Recent" meaning "not past EOL", so not including 6.8. I don't recall the details in 6.8 although I don't recall ever seeing anyone else doing periodic synced flushes. If it were me I'd focus on upgrading to 7.x instead.
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.