Docker - influxdb high cpu use

InfluxDB 2
, ,
1

Hi, I am trying to create Candlestick from data I enter from binance, they are 1m timeframe candles.
When querying the data it does not take more than 0.07s
But when I implement some of the functions that I found on this website for example this code, the cpu usage goes way up

data = from(bucket: "crypto_app")
  |> range(start: v.timeRangeStart, stop: v.timeRangeStop)
  |> filter(fn: (r) => r["_measurement"] == "candles")
  |> filter(fn: (r) => r["exchange"] == "Binance")
  |> filter(fn: (r) => r["symbol"] == "BTC-USDT")
  |> window(every: 5m)
  |> reduce(fn: (r, accumulator) => ({
  
      indexLow:
        if (r._field=="low") then 
          accumulator.indexLow+1 
        else
        accumulator.indexLow,
      indexOpen: 
      if (r._field=="open") then 
       accumulator.indexOpen+1 
       else 
       accumulator.indexOpen,
        open: 
      if (r._field=="open") then 
        if (accumulator.indexOpen==0) then 
           float(v:r._value )
        else 
          accumulator.open
      else
        accumulator.open  
    ,
    
    
      high:
       if (r._field=="high") then  
          if(r._value>accumulator.high ) then
            float(v:r._value )
          else
            accumulator.high 
      else 
        accumulator.high
   ,
    low: 
      if (r._field=="low") then

          if(r._value<accumulator.low or accumulator.indexLow==0.0) then
            float(v:r._value )
          else
           accumulator.low 
      else 
        accumulator.low,

             close: 
       if (r._field=="close") then 
          float(v:r._value )
      else 
        accumulator.close,
             volume: 
        if (r._field=="volume") then
          float(v:r._value )+accumulator.volume 
          else
           accumulator.volume
        
  
    }),
    identity: {indexLow:0,indexOpen:0,open: 0.0,high: 0.0,low: 0.0,close: 0.0,volume: 0.0})
    |> drop(columns: ["indexOpen","indexLow"])
    |> group(columns:["symbol"])
    |> yield(name: "candle")

Another code I found

data = from(bucket:"crypto_app")
  |> range(start: v.timeRangeStart, stop: v.timeRangeStop)
  |> filter(fn: (r) => r["_measurement"] == "candles")
  |> filter(fn: (r) => r["symbol"] == "BTC-USDT")
 

open = data |> filter(fn: (r) => r._field == "open") |> aggregateWindow(every: 5m, fn: first) |> drop(columns: ["_start", "_stop", "_measurement", "_field"])
close = data |> filter(fn: (r) => r._field == "close") |> aggregateWindow(every: 5m, fn: last) |> drop(columns: ["_start", "_stop", "_measurement", "_field"])
low = data |> filter(fn: (r) => r._field == "low") |> aggregateWindow(every: 5m, fn: min) |> drop(columns: ["_start", "_stop", "_measurement", "_field"])
high = data |> filter(fn: (r) => r._field == "high") |> aggregateWindow(every: 5m, fn: max) |> drop(columns: ["_start", "_stop", "_measurement", "_field"])


OC = join(tables: {o: open, c: close}, on: ["_time", "symbol"])
HL = join(tables: {h: high, l: low}, on: ["_time", "symbol"])
OCLH = join(tables: {OC: OC, HL: HL}, on: ["_time", "symbol"])

|> group(columns:["symbol"])
|> rename(columns: {_value_c: "c", _value_h:"h", _value_l:"l", _value_o:"o"})
|> yield(name: "mean")

Screenshot_32
Screenshot_321613×698 83.1 KB

In this way I entered the information using a python script from the csv in binance collection
point = (
dot(“candles”)
.tag(“exchange”, “Binance”)
.tag(“symbol”, “BTC-USDT”)
.tag(“interval”, “1m”)
.tag(“topic”, “crypto/candles/Binance/BTC-USDT/1m”)
.field(“start”, row[0].astype(‘int64’))
.field(“stop”, row[6].astype(‘int64’))
.field(“open”, row[1])
.field(“high”, row[2])
.field(“minimum”, row[3])
.field(“close”, row[4])
.field(“volume”, row[5])
.time(row[0].astype(‘int64’), “ms”),
)
1722470400000,64628.01000000,64670.01000000,64601.00000000,64634.01000000,36.77990000,1722470459999,2377477.74237530,3399,13.19798000,853026.66151230,0
sometimes it takes more than 50sec or never ends.
I have no idea what I am doing wrong

2

@Jose_Martinez,
Yah unfortunately that isn’t surprising to me. How many points are you querying?
That flux is taking that long to run. I would suggest doing this work with python instead.
You could do some of this processing before it hits the db with the exec plugins.