elasticsearch-py and multiprocessing

Question

What is the correct way to use elasticsearch-py in multiprocessing script? Should I create a new client object before start processes and use that object or should I create a new object inside each of the processes. The 2nd one gives me an an error with connection issues from elasticsearch

Thanks Kiran

The second one should work (works for me). Could you show traceback / code sample? — Oleksiy, Mar 27 '15 at 04:47

score 3 · Answer 1 · answered Nov 12 '18 at 03:21

It seems the first method works for me, when I declare the client object as a global variable.

from multiprocessing import Pool
from elasticsearch import Elasticsearch
import time


def task(body):
    result = es.index(index='test', doc_type='test', body=body)
    return result


def main():
    pool = Pool(processes=MAX_CONNECTS)
    result = []
    for x in range(10):
        result.append(pool.apply_async(task, ({'id': x},)))
    time.sleep(1)
    for rs in result:
        print(rs.get())


if __name__ == "__main__":
    MAX_CONNECTS = 5
    es = Elasticsearch(hosts="localhost", maxsize=MAX_CONNECTS)
    main()

The output looks like

{'_index': 'test', '_type': 'test', '_id': 'xEjqBWcB9xsUYKqz-P6U', '_version': 1, 'result': 'created', '_shards': {'total': 2, 'successful': 1, 'failed': 0}, '_seq_no': 1, '_primary_term': 1}
{'_index': 'test', '_type': 'test', '_id': 'w0jqBWcB9xsUYKqz-P6U', '_version': 1, 'result': 'created', '_shards': {'total': 2, 'successful': 1, 'failed': 0}, '_seq_no': 0, '_primary_term': 1}
{'_index': 'test', '_type': 'test', '_id': 'x0jqBWcB9xsUYKqz-P6X', '_version': 1, 'result': 'created', '_shards': {'total': 2, 'successful': 1, 'failed': 0}, '_seq_no': 4, '_primary_term': 1}
{'_index': 'test', '_type': 'test', '_id': 'xkjqBWcB9xsUYKqz-P6X', '_version': 1, 'result': 'created', '_shards': {'total': 2, 'successful': 1, 'failed': 0}, '_seq_no': 3, '_primary_term': 1}
{'_index': 'test', '_type': 'test', '_id': 'xUjqBWcB9xsUYKqz-P6W', '_version': 1, 'result': 'created', '_shards': {'total': 2, 'successful': 1, 'failed': 0}, '_seq_no': 2, '_primary_term': 1}
{'_index': 'test', '_type': 'test', '_id': 'yEjqBWcB9xsUYKqz-P66', '_version': 1, 'result': 'created', '_shards': {'total': 2, 'successful': 1, 'failed': 0}, '_seq_no': 4, '_primary_term': 1}
{'_index': 'test', '_type': 'test', '_id': 'ykjqBWcB9xsUYKqz-P7I', '_version': 1, 'result': 'created', '_shards': {'total': 2, 'successful': 1, 'failed': 0}, '_seq_no': 2, '_primary_term': 1}
{'_index': 'test', '_type': 'test', '_id': 'yUjqBWcB9xsUYKqz-P7I', '_version': 1, 'result': 'created', '_shards': {'total': 2, 'successful': 1, 'failed': 0}, '_seq_no': 3, '_primary_term': 1}
{'_index': 'test', '_type': 'test', '_id': 'y0jqBWcB9xsUYKqz-P7P', '_version': 1, 'result': 'created', '_shards': {'total': 2, 'successful': 1, 'failed': 0}, '_seq_no': 4, '_primary_term': 1}
{'_index': 'test', '_type': 'test', '_id': 'zEjqBWcB9xsUYKqz-P7V', '_version': 1, 'result': 'created', '_shards': {'total': 2, 'successful': 1, 'failed': 0}, '_seq_no': 5, '_primary_term': 1}

score 0 · Answer 2 · answered Dec 28 '16 at 14:41

0

The recommended way is to create a unique client object and you can increase the number of simultaneous thread using the maxsize (10 by default).

es = Elasticsearch( "host1", maxsize=25)

Source

answered Dec 28 '16 at 14:41

ML_TN

727
6
16

1

Wrong: In the link you have posted above "Since we use persistent connections throughout the client it **means that the client doesn’t tolerate fork very well**. If your application calls for multiple processes make sure you create a fresh client after call to fork. Note that Python’s multiprocessing module uses fork to create new processes on POSIX systems." – Antoine Sep 15 '20 at 15:36

elasticsearch-py and multiprocessing

2 Answers2