Mongoimport fails with broken pipe when the file is large

Working with Data Developer Tools
#1

Hi,

I’m trying to import a dataset, which is in a large file (21.1GB json), which has a list of documents. I could import around 75% (15.9GB) of that file and then it gives “error inserting documents: write tcp a.b.c.d:e: write broken pipe”

I tried this 3 times and each time this happened. The command I’m using is as below (connecting to mongos instance);

mongoimport -d dbname -c collectionName --host hostname --file filename -j 4 --batchSize=200 --jsonArray

Has anybody faced the same issue and any recommendations?

Thanks

Failed: connection(cluster0-shard-00-02.vwxk3a.mongodb.net:27016[-16]) unable to write wire message to network: write tcp : write: broken pipe
#2

There might be a message related to the error in the server logs. It might provide more details.

Here is some related information at: https://jira.mongodb.org/browse/TOOLS-379

#3

Hi,

Thanks for the reply… However, if the document is bigger than 16MB, MongoDB skips importing that particular document and it’s printed on the command line… But this is something else I guess… Will post if I find out the reason…

#4

Hi @Laksheen_Mendis,

What does mongoimport --version report and what specific version of MongoDB server are you importing into? Also, what type of deployment do you have (standalone, replica set, or sharded cluster)?

Finally: how large are your documents on average (you can check imported documents via db.collectionName.stats().avgObjSize)? If you have large documents you may want to try further reducing the --batchSize value.

Regards,
Stennie

1 Like