We're on Elasticsearch 7.9. We have a percolate index that hosts about 120 queries. The index has 5 shards, each of which has two replicas.
We started to encounter performance issues after adding around 10 queries that look like this:
query/query at main · jessechen789/query · GitHub (the text is too long to paste in this post)
When traffic is around 400 reqs/s, Stack monitoring shows avg search time from 100ms to 200ms and request time around 100,000ms.
Here is an example request:
{
"query":{
"bool":{
"filter":[
{
"percolate":{
"field":"query",
"string1":"They created an indoor-version of the Google Street View 360-degree camera system to capture gallery images by pushing the camera trolley through a museum. It also used professional panoramic heads Clauss Rodeon VR Head Hd And Clauss VR Head ST to take high-resolution photos of the artworks within a gallery. This technology allowed the excellent attention to detail and the highest image resolution. Each partner museum selected one artwork to be captured at ultra-high resolution with approximately 1,000 times more detail than the average digital camera.[2] The largest image, Alexander Andreyevich Ivanovs The Apparition of Christ to the People, is over 12 gigapixels. To maximize image quality, the team coordinated with partner museums lighting technicians and photography teams. For example, at the Tate Britain, they collaborated to capture a gigapixel image of No Woman No Cry in both natural light and in the dark. The Tate suggested this method to capture the painting's hidden phosphorescent image, which glows in the dark. The Google camera team had to adapt their method and keep the camera shutter open for 8 seconds in the dark to capture a distinct enough image. Now, unlike at the Tate, from the site, one can view the painting in both light settings.",
"number1":1605,
"string2":"teacher"
}
}
]
}
}
}
Any tips for performance optimization will be appreciated. Thanks!