Mongodb dump size too smaller than it was in the database

I want migrate old Mongodb2.6.8 database to new Mongodb4.4. So I tried mongodump to backup below "okapp" database.

[okapp@centos101 bin]$ ./bin/mongo localhost:27018
MongoDB shell version: 2.6.8
connecting to: localhost:27001/test
> show dbs
admin    (empty)
okapp    6.497GB
ca423    0.031GB
local    0.031GB
>

As you can see okapp size is 6.497GB. But after i run below:

./mongodump -h localhost:27001 -d okapp

And I will get a the dump:

2021-11-02T15:47:14.550+0100 DATABASE: okapp to     dump/okapp
2021-11-02T15:47:14.550+0100    okapp.system.indexes to dump/okapp/system.indexes.bson
2021-11-02T15:47:14.551+0100             5 documents
2021-11-02T15:47:14.551+0100    okapp.findjobsee to dump/okapp/findjobsee.bson
2021-11-02T15:47:14.551+0100             10 documents
2021-11-02T15:47:14.551+0100    Metadata for okapp.findjobsee to dump/okapp/findjobsee.metadata.json
2021-11-02T15:47:14.552+0100    okapp.findtasksee to dump/okapp/findtasksee.bson
2021-11-02T15:47:14.552+0100             0 documents
2021-11-02T15:47:14.552+0100    Metadata for okapp.findtasksee to dump/okapp/findtasksee.metadata.json
2021-11-02T15:47:14.552+0100    okapp.findresultsee to dump/okapp/findresultsee.bson
2021-11-02T15:47:14.614+0100             2937 documents
2021-11-02T15:47:14.614+0100    Metadata for okapp.findresultsee to dump/okapp/findresultsee.metadata.json
2021-11-02T15:47:14.615+0100    okapp.findjobarchivesee to dump/okapp/findjobarchivesee.bson
2021-11-02T15:47:14.618+0100             751 documents
2021-11-02T15:47:14.618+0100    Metadata for okapp.findjobarchivesee to dump/okapp/findjobarchivesee.metadata.json

[okapp@centos101 somedir]$ cd ./mongodb-linux-x86_64-2.6.8/dump/okapp
[okapp@centos101 somedir]$ du -sh  okapp
7.7MB okapp
[okapp@centos101 somedir]$ cd ./mongodb-linux-x86_64-2.6.8/dump/okapp
[okapp@centos101 okapp]$ du -sh *
316K    findalljobarchivesee.bson
4.0K    findalljobarchivesee.metadata.json
4.0K    findalljobsee.bson
4.0K    findalljobsee.metadata.json
7.4M    findallresultsee.bson
4.0K    findallresultsee.metadata.json
0       findalltasksee.bson
4.0K    findalltasksee.metadata.json
4.0K    system.indexes.bson
[okapp@centos101 okapp]$ 
[okapp@centos101 okapp]$

Note: i've checked the collections in the Mongodb

> db.findjobarchivesee.count()
751
> db.findjobsee.count()
10
> db.findresultsee.count()
2937
> db.findtasksee.count()
0
> db.system.indexes.count()
5

Anyone know why the size shrink too much? It was 6.497GB in database, but it just a 7.7 MB after mongodump.

Solution

I think there are a few things that contribute to the difference:

MMAP does not compact. The size reported by show dbs includes empty space in the files, which may be due to documents being deleted.
MongoDB with MMAP pads each document so there is free space between them to allow for documents to grow. This padding will not be included with the documents when exported.
mongodump does not export index data. It exports the information that mongorestore will need in order to recreate the index at restore time.