I have the following setup. A java program, using ES6.0 high/low level libraries, it:
queries an ES2.2 cluster, using scrolling
creates new docs out of the 2.2 ones
indexes them into a ES6.0 cluster
Of course, writing to the 6.0 cluster is no issue. Reading off the 2.2 index mostly works too, but it fails when trying to clear the scroll:
clearScrollRequest.addScrollId(scrollId);
Unable to parse response body for Response{requestLine=DELETE /_search/scroll HTTP/1.1, host=http://...:9200, response=HTTP/1.1 200 OK}
If I comment out this code it runs fine.
Now, I am tailing the 2.X index, so I run the scrolling query every min (long story...we are transitioning to 6.0 and this was a pragmatic way to keep the current setup working). Does it sound ok to run this without closing the scroll (remember, once a min...), or should I expect some issue ??
yes, I am aware about it not being officially supported, my question was more about possible bad side effects of not clearing the scroll....I'l try with your suggestion, thanks
It will hold the segments on disk as they are until the scroll is cleaned. Which means that even though some segment merges are happening, the "old" ones won't be removed until all locks are gone.
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.