I have two more or less identical servers in my cluster that each have 3 disks in the data_path each with 1.5TB.
Recently one of the server has started running out of disk on one of the three disks (the first in the list if it matters).
Over all there is plenty of disk space so the high water mark is never reached ES just suddenly runs out of space on one disk then and shuts down.
the disk watermarks are per-node things: if one node starts to run out of space then it will hit its watermark, even if there's plenty of space overall.
when a node runs out of space it doesn't shut down.
Can you clarify what you mean exactly, with logs and so on? We only check disk space every 30s or so, so if you write a huge amount of data in 30s and/or the watermarks are set very close then it is possible to go from below the high watermark to a full disk before any protection mechanisms kick in.
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.