Closed jaserud closed 3 years ago
max_buckets
configs as well For aggregations, track_total_hits
is need if we want totalHits
to be correct from search:
However, we will still get an error if aggregation returns more buckets than max_buckets
@jaserud thanks - i guess song-search is not using any buckets explicitly, but should we set this to be the same as totalhits
anyways?
@rosibaj The search.max_bucket
is actually a cluster wide config, and I'm not sure how it affects the cluster so I think it might need a bit more thinking before setting it to a default based on song-search's totalHits. The default for this value is supposed to be 65,536 actually.
Currently song-search has no bucket aggs. Workflow search has one on the run.states (bucket runs by states) and there are only a dozen or so states.
So my two cents, I think the current rdpc-gateway shouldn't have any issues with the default max_buckets so we can leave it their :thinking:.
The fix was released in 2.6.0 which has already been released to prod
Describe the bug
If index has more then 10000 documents, the
info.totalHits
only returns 10000 as value even if there are actaully more than that.Example using it on an index where files_centric index has more than 10000 documents:
This is only an issue if file_centric (or analysis_centric) has more than 10000 documents.
Steps To Reproduce
Steps to reproduce the behaviour:
from
> 10000 and query for info.totalHitsExpected behaviour
Should return correct number of total hits. The bug is most likely due to elastic search and its 10000 query limit so the fix might be their.
Note: this could be an issue in workflow-search too, but need to check.