Hi there,
I'm running Kibana 6.5 and I have a data table visualisation where I'm showing the 95th percentile of a field, however the value is too inconsistent due to the approximation calculation for percentiles. I've seen that you can set the compression value for the tdigest approximation algorithm in an Elasticsearch query to improve the accuracy at the cost of more memory usage.
I'm looking to see the following included in the visualisation's query to Elasticsearch:
{
"aggs": {
...
"3": {
"percentiles": {
"field": "x",
"tdigest": {
"compression": 1000
},
"percents": [
95
],
"keyed": false
}
}
}
...
}
On a guess, I tried adding the tdigest setting to the visState JSON for the visualisation in a way that seemed like it would correlate to the desired Elasticsearch query:
{
...
"aggs": [
...
{
"id": "3",
"enabled": true,
"type": "percentiles",
"schema": "metric",
"params": {
"field": "x",
"tdigest": {
"compression": 1000
},
"percents": [
95
],
"customLabel": "X Value"
}
}
]
}
However, when I inspected the query it sent to Elastic, the tdigest parameters weren't included.
{
"aggs": {
...
"3": {
"percentiles": {
"field": "x",
"percents": [
95
],
"keyed": false
}
}
}
...
}
How can I set the tdigest compression value for a visualisation? Is this even possible to do?
Thanks