i have 2 daily indices that are about 30-40Gb every day and even spread by 6 shards are big. When i have to restart one node, they take a long time to recover.
So i want to rotate then not by day, but by every 6 hours ... how can i do that? %{+YYYY.MM.dd} is daily, %{+YYYY.MM.dd-HH} is hour, how can i do a %{YYYY.MM.dd-H%6} ?
maybe i could use elasticsearch _rollover, but how to use it? i found no info about adding this to the templates
Sorry, actually is 3 shards and 1 replica.
the problem i'm trying to solve is the cluster recovery time, with 2000 indices and around 13k shards takes between 1h to 3 hours to recover... and bigger indices are the ones that take longer.
by breaking those big indexes in 4 smaller shards, i hope that ES will recover then faster, as at least 3 of then didn't change and full recovering the last one would be faster as it is much smaller
Having a very large number of shards to move around can also have a negative impact on recovery. 6GB is generally not very large, so I would probably not go smaller than that. Recovery might actually be helped by increasing the size rather than decreasing it. The only way to find out for sure is probably to test it though.
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.