CODEWITHSUNDEEP
node.jselasticsearch
ryu
ryu

Reputation: 661

Elastic search bulk index timeout err! Error: Request Timeout after 30000ms

For recently, I want to scroll through the old index data to new monthly-based indices. The stored data begin from 2015/07 until now. and it is almost 30,000 records for every month. Follow the scroll and bulk methods provided in 2.2 API, I finish the code as follows.

file main.coffee

logger = require 'graceful-logger'
elasticsearch = require 'elasticsearch'
setMonthlyIndices = require './es-test-promise'
client = new elasticsearch.Client
  host:
    host: 'localhost'
    port: 9200
    protocol: 'http'
setMonthlyIndices client, 'es_test_messages', 'talk_messages_v2', 'messages', 2015, 6

file es-test-promise.coffee

logger = require 'graceful-logger'
elasticsearch = require 'elasticsearch'
config = require 'config'

setIndice = (client, prefix, index, type, year, month) ->
  allDocs = []
  count = 0

  prevYear = year + ''
  # with leading '0' for month less than 10
  prevMonth = ("0" + month).slice(-2)
  nextDate = new Date(year, month)
  nextYear = nextDate.getFullYear().toString()
  nextMonth = ("0" + (nextDate.getMonth()+1)).slice(-2)

  minDate = "#{prevYear}-#{prevMonth}-01"
  maxDate = "#{nextYear}-#{nextMonth}-01"

  indice_name = "#{prefix}_#{prevYear}_#{prevMonth}"

  q =
    filtered:
      filter:
        range:
          createdAt:
            gte: minDate
            lt: maxDate
            format: "yyyy-MM-dd"

  client.search
    index: index
    type: type
    scroll: '1m'
    body:
      query: q
    sort: ['_doc']
    size: 1000
  , callback = (err, response) ->
    console.log "indice_name 1", indice_name
    return logger.err err.stack if err
    return unless response.hits?.total

    allDocs = []

    response.hits.hits.forEach (hit)->
      action =
        index:
          _id: hit._id
      allDocs.push(action)
      allDocs.push(hit._source)

    count = count + allDocs.length

    client.bulk
      index: indice_name
      type: type
      body: allDocs
    , (err, resp) ->
      console.log "indice_name 2", indice_name
      return logger.err err.stack if err

      if response.hits.total *2 !=  count
        client.scroll
          scrollId: response._scroll_id
          scroll: '1m'
        , callback
      else
        logger.info "Finish indicing #{indice_name}"

setMonthlyIndices = (client, prefix, index, type, year, month) ->
  current = new Date()
  currentYear = current.getFullYear()
  currentMonth = current.getMonth() + 1

  processYear = year or currentYear
  processMonth = month or 0

  processDate = new Date(processYear, processMonth)
  currentDate = new Date(currentYear, currentMonth)

  processDate = new Date(2015, 6)
  currentDate = new Date(2015, 9)

  while processDate <= currentDate
    year = processDate.getFullYear()
    month = processDate.getMonth() + 1
    setIndice(client, prefix, index, type, year, month)
    processDate.setMonth(processDate.getMonth() + 1)

module.exports = setMonthlyIndices

I am wondering whether it is due to open too many client request, because in file es-test-promise.coffee, all these search request is running simultaneously. This is just a guess, and then I have also tried to implement with promise to make sure the request could be executed one by one. Finally， I can't figure it out and give up.

Do you have any suggestion, I think it should be the source issues , but I don't know where to check...

Upvotes: 1

Views: 6223

Answers (3)

illusionx
illusionx

Reputation: 3845

Adding more clarity in case if someone needs, the following should help

  1. Increasing timeout, max_retries and retry on timeout in Elasticsearch connector
from elasticsearch import Elasticsearch
es_mi_prod = Elasticsearch(
    ["host"],
    scheme="https",
    port=443,
    # ssl_context=context,
    http_auth=("user_name", "pass"),
    timeout=120,
    max_retries=10,
    retry_on_timeout=True
)
  1. Decreasing the chunk size and increasing the request timeout in helpers bulk request
helpers.bulk(
   connector, 
   generator_function, 
   chunk_size=500, 
   request_timeout=120
)

Reference - ES Python Docs

Upvotes: 0

Shudipta Sharma
Shudipta Sharma

Reputation: 5852

In Elasticsearch 7.x the default timeout is 1m (one minute) here.

The official go client of Elasticsearch go-elasticsearch has a way to set this value.

// WithTimeout - explicit operation timeout.
//
func (f Bulk) WithTimeout(v time.Duration) func(*BulkRequest) {
    return func(r *BulkRequest) {
        r.Timeout = v
    }
}


esutil.BulkIndexerConfig {
    // ...
    Timeout: timeout,
}

Upvotes: 0

bhushanRJ
bhushanRJ

Reputation: 164

Just put the requestTimeout to your config.

e.g :

new elasticsearch.Client({host:"localhost", requestTimeout : Infinity});

You can replace Infinity by your desired limit in 'ms' .

Upvotes: 3

Related Questions