Since we upgraded to 1.4.0, deleted indices in our time-series index set
keep coming back right after deletion. So whenever we drop an expired index
(usually as midnight rolls), it gets deleted and removed from the alias it
was under. But about half the time it comes back as an empty index.
As you can see from the marvel screenshot below (read from the bottom to
the top).
That's unlikely to be a bug, the only time ES will recreate an index is if
it finds dangling data.
Are your indexes created automatically? How is your data sent to ES? Is it
possible that there is some data that is slower reaching ES than others and
so the time difference causes this to happen?
Since we upgraded to 1.4.0, deleted indices in our time-series index set
keep coming back right after deletion. So whenever we drop an expired index
(usually as midnight rolls), it gets deleted and removed from the alias it
was under. But about half the time it comes back as an empty index.
As you can see from the marvel screenshot below (read from the bottom to
the top).
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.