Solr import/export

README.md

Solr import/export

You need to move from one solr instance to another and can't be bothered with mismatching versions or whatever? These two scripts will help you :)

First you need to create a new core in the target instance. You may want to use the schema/configset from the originating instance though, as the default schema might not be ideal.

Im my scenario I moved from Solr 5.5.5 to Solr 7.4. Therefore I had to (at least) update the solrconfig.xml, where the lucene version is specified. The exact version you need can be found in the default configset ([solr_root]/server/solr/configsets/...)

The first script downloads and entire core into one file (one json document per line).

python download.py 172.0.0.1 8983 my_core my_core.json

The second one uploads it to a server.

python uploadSolr.py 172.0.0.1 8983 my_core my_core.json

IMPORTANT

The download script is extremely stupid. Once it's done, it will continue requesting the server. Fortunately tqdm will give you a rough estimate for how long it may take to finish, so you know how long your coffe break can be. There will be no harm if it runs for a while (except heating up your server). Once the progress bar is full, kill the script with Ctrl+C.

FYI: Downloading 2.5G (380k docs) took 5 minutes, uploading them again took just over an hour.

Requirements to install:

• pysolr
• requests
• tqdm

You could remove tqdm from the code. It's for a nice progress bar. requests could also be removed, but I'm to lazy to update the download scrip to also use pysolr (or vice versa).

download_solr.py

import requests
import json
import argparse
from tqdm import tqdm

parser = argparse.ArgumentParser(description='Download all documents from a solr core')
parser.add_argument('HOST', help='Host of solr server, e.g. 172.0.0.1 for localhost')
parser.add_argument('PORT', help='Port of solr server, e.g. 8983 for default port')
parser.add_argument('CORE', help='Core to fetch from server')
parser.add_argument('TARGET', help='target file to write to (one line = one doc as json)')
parser.add_argument('--rows', help='number of rows per request', default=100)
args = parser.parse_args()

target_dir = ''
solr_url = args.HOST
solr_port = args.PORT
solr_core = args.CORE

r = requests.get('http://{}:{}/solr/{}/select?q=*%3A*&rows=1&sort=id+asc&wt=json&cursorMark=*'.format(
    solr_url, solr_port, solr_core))
res = r.json()
num_results = res['response']['numFound']

with tqdm(total=num_results) as pbar, open(args.TARGET, 'w') as f:
    cursor_mark = '*'
    while True:
        r = requests.get('http://{}:{}/solr/{}/select?q=*%3A*&rows={}&sort=id+asc&wt=json&cursorMark={}'.format(
            solr_url, solr_port, solr_core, args.rows, cursor_mark))
        res = r.json()

        for doc in res['response']['docs']:
            f.write(json.dumps(doc) + '\n')
        
        # alternative log to tqdm:
        # print('Got {} docs | {}/{}  ({:.2f}%)'.format(len(res['response']['docs']),
        #                                              cnt, res['response']['numFound'],
        #                                              cnt / res['response']['numFound']))
        pbar.update(len(res['response']['docs']))
        cursor_mark = res['nextCursorMark']

upload_solr.py

import pysolr
import json
import argparse
from tqdm import tqdm

parser = argparse.ArgumentParser(description='Download all documents from a solr core')
parser.add_argument('HOST', help='Host of solr server, e.g. 172.0.0.1 for localhost')
parser.add_argument('PORT', help='Port of solr server, e.g. 8983 for default port')
parser.add_argument('CORE', help='Core to write to on server')
parser.add_argument('SOURCE', help='target file to read from (one line = one doc as json)')
parser.add_argument('--rows', help='number of rows per request', default=100)
args = parser.parse_args()

target_dir = ''
solr_url = args.HOST
solr_port = args.PORT
solr_core = args.CORE

client = pysolr.Solr('http://{}:{}/solr/{}'.format(solr_url, solr_port, solr_core))
num_docs = sum(1 for line in open(args.SOURCE))

with tqdm(total=num_docs) as pbar, open(args.SOURCE, 'r') as f:
    buffer = []
    for line in f:
        doc = json.loads(line)
        del doc['_version_']
        buffer.append(doc)
        if len(buffer) >= args.rows:
            client.add(buffer)
            pbar.update(len(buffer))
            buffer = []
    client.add(buffer)
    pbar.update(len(buffer))

Hi @karcool ,

it might help to print the contents of res and go from there. The cursorMark is essential for paginating through the entire query set and should work. I'm using this in Solr 9.3 without problems and I can't imagine something changed in 9.4 (assuming you are running the latest Solr version). If res does not yield a helpful error message, you may have a look at the solr logs, check response headers, and study the changelog (I couldn't find anything related past version 8.x).

Good luck!

Amazing, thank you so much.