I am migrating several hundred million tweets of the format {'id_str': , 'created_at': , 'text': } from text files into MongoDB using pymongo. A collection is created for each user to store his/her tweets. The insertion method I am using is insert_many(). It often runs into BulkWriteError.
Traceback (most recent call last):
File "pipeline.py", line 105, in <module>
timeline_db, meta_db, negative_db, log_col, dir_path)
File "/media/haitao/Storage/twitter_pipeline/migrate_old.py", line 134, in migrate_dir
timeline_db[user_id].insert_many(utility.temporal_sort(statuses))
File "/home/haitao/anaconda3/envs/py27/lib/python2.7/site-packages/pymongo/collection.py", line 711, in insert_many
blk.execute(self.write_concern.document)
File "/home/haitao/anaconda3/envs/py27/lib/python2.7/site-packages/pymongo/bulk.py", line 493, in execute
return self.execute_command(sock_info, generator, write_concern)
File "/home/haitao/anaconda3/envs/py27/lib/python2.7/site-packages/pymongo/bulk.py", line 331, in execute_command
raise BulkWriteError(full_result)
pymongo.errors.BulkWriteError: batch op errors occurred
This error seems to occur when there are duplicates of keys, which should not be the case for here. Are there other things that I can check to solve this issue?
Thanks in advance!
sorry for the delay.
1) I replicated the error. The following is close to the end of the mongod.log.
I - [ftdc] Assertion: 13538:couldn't open [/proc/5774/stat] errno:24 Too many open
files
W FTDC [ftdc] Uncaught exception in 'Location13538: couldn't open [/proc/5774/stat]
errno:24 Too many open files' in full-time diagnostic data capture subsystem. Shutting down
the full-time diagnostic data capture subsystem.
E STORAGE [conn2] WiredTiger (24) [1491020356:127332][5774:0x7f6f30e9d700], WT_SESSION
.create: /var/lib/mongodb/: directory-sync: open: Too many open files
I COMMAND [conn2] command timeline_db.231731006 command: insert { insert: "231731006",
ordered: true, documents: 1000 } ninserted:0 keyUpdates:0 writeConflicts:0 exception: 24:
Too many open files code:8 numYields:0 reslen:123 locks:{ Global: { acquireCount: { r: 2,
w: 2 } }, Database: { acquireCount: { w: 1, W: 1 } }, Collection: { acquireCount: { w: 1,
W: 1 } } } protocol:op_query 511ms```
2) Yes, only one instance of MongoClient() is passed around.
3) No multi-processing was run.
After I posted the initial question, I started to use insert_one() which explicitly raised open file limit error. I changed the design of the database (mainly, reduced the number of collections) and solved the issue about open file limit. I am not sure, but the log seems to suggest that the actual cause of the BulkWriteError is also open file limit.
