I unfortunately have to say that this did not solve the issue in the end.
As mentioned in https://github.com/elastic/elasticsearch/issues/24390, the problem is that ES iterates over all file stores on startup. So I do not have an idea how to fix this apart from waiting for release 6.0.0.
If there is any known workaround please let me know.
Avoiding stale nfs mounts is NOT possible in our environment.
Finally, it seems that we could at least avoid the hangup behaviour at client side.
We did this by modifiying the static initializer in org.elasticsearch.env.Environment. Instead of iterating on all filestores it simply creates an array containing only one filestore - the one behind the current directory. We assume that on client side the file store has no relevance as there is no data to be stored.
Just for curiosity I would like to know, what for ES needs all the file stores. E.g, when the data directory is located elsewhere than the installation directory?
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.