5.x Percolator memory/cache issue


#1

Hi,

While testing 5.x percolator we ran into serious performance problems when percolating of a single document could take up to 30 seconds instead of usual 200 ms. Our test is sending documents to percolate in infinite loop. Once ES service is restarted the percolating is taking around 200 ms on average; after a few hours (could be an hour, could be 3-4) it starts to slow down and becomes very slow very quickly. Restarting ES service or executing /_cache/clear command fixes this problem. Please advice on what we are doing wrong.

Thank you,

Yuri


(Martijn Van Groningen) #2

@ymetelkin Are you by any chance percolator document that have a nested field? If so you may have ran into this bug [1] which has been fixed [2]

1: https://github.com/elastic/elasticsearch/issues/24108
2: https://github.com/elastic/elasticsearch/pull/24115


#3

We building ES from repositories, the version in question is 5.3.1. What version is it fixed in? And yes, we are using nested queries.


(Martijn Van Groningen) #4

The fix is included in the upcoming 5.4 release.


#5

Thank you. Any idea when it is going to be available?


(Martijn Van Groningen) #6

I can't provide a specific date, but it is planned to be released soon. I would say within 2 weeks.


(system) #7

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.