Update on multiple records inside collection

Important note: This article is in relation to online MongoDB course. For more information about the course and other posts describing its content please check my main page here: M101P: MongoDB for Developers course. 

By default an update function with 2 arguments updates a SINGLE RANDOM (unspecified) document. To update multiple documents we need to provide the 3th option: multi.

The {} as first argument means match every document.

 
> db.arrays.find()
{ "_id" : ObjectId("52ba1c78a83c1ee5e6c903a1"), "a" : 1, "tab" : [  11,  22,  "hello" ] }
{ "_id" : ObjectId("52ba1e4cc1207a4a1fdeb3ac"), "a" : 2, "tab" : { "2" : "london" } }
{ "_id" : ObjectId("52ba1e99c1207a4a1fdeb3ad"), "a" : 3, "mystr" : "poland" }

> db.arrays.update( { }, { $inc : { "a" : 1 } }, { multi: true} )
>
> db.arrays.find()
{ "_id" : ObjectId("52ba1c78a83c1ee5e6c903a1"), "a" : 2, "tab" : [  11,  22,  "hello" ] }
{ "_id" : ObjectId("52ba1e4cc1207a4a1fdeb3ac"), "a" : 3, "tab" : { "2" : "london" } }
{ "_id" : ObjectId("52ba1e99c1207a4a1fdeb3ad"), "a" : 4, "mystr" : "poland" }

IMPORTANT: MongoDB doesn't offer isolated transactions across multiple documents. It grantees a single document update is atomic. No other concurrent updates/writes can modify the same document.

In regards to the test above it means that there is no grantee that all the updates will be executed in one go.

It is possible that the db engine thread that does the update will be stopped, de-scheduled or its execution will be yielded (will be give) to another db engine thread that modifies the same document collection (or its subset). In such a case the updates on the documents may happen in different orders and one can override the other. Only a single document update is atomic.