Scroll analysis at node stats level

I was querying the nodes API in an ES cluster and saw the following output related to scroll contexts.

"scroll_total": 1546241427,
"scroll_time": "103866.7d",
"scroll_time_in_millis": 8974085281558,
"scroll_current": 1396,

Reading throug the guide I got that the scroll_total referes to done total scroll operations, scroll_current are current scroll contexts open. And scroll time is the value that is most curious to me, 103866.7d, those are 103866 days, which are more time than the time my company has being using ES. My question is:

Is the scroll time the time that the scroll contexts will be left open even if they are not being used anymore?

How to calc the impact of this scroll contexts on the node resources, mem and cpu usage? I know for sure there isn't a way to do that, but we are enthusiast people :smiley:

Thanks in advance for all of your answers.

Yes, and it counts each scroll separately so if you have 10 scrolls open then after an hour that counts as 10 hours of scroll time. I think you should consider it a bug in your usage of Elasticsearch if you're leaving scroll contexts open after they're not being used any more.

Open scrolls may not take up significant resources but they block the cluster from doing certain things, e.g. completely relocating shards, that will cause you problems if not addressed.

1 Like

Thanks for your answer @DavidTurner.

1 Like

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