Cosmos Db save call is failing intermittently with "PreConditionFail"

[Karthickramk]

Describe the bug
Cosmos Db save call is failing intermittently with "PreConditionFail". Even though there is no other call is invoked for that particular document at that particular time. We have a retry for this exception. During retry, we reload the data again from database and do the operation. But still we see the same error.
I am using 2.6.6 SDK.

To Reproduce
There is no concrete reproducing step as this very intermittent issue.

Expected behavior
The save call should be success.

Actual behavior
Db call fails with "PreConditionFail"

Environment summary
SDK Version:2.6.6
Java JDK version:openjdk 11.0.4.0.101 2019-10-15 LTS
OS Version (e.g. Windows, Linux, MacOSX) Azure(AKS)

Additional context

[kushagraThapar]

@Karthickramk - "PreConditionFail" means the call is failing because of etag mismatch.
Seems like the etag you have locally on the document is different from what is stored in the database.
Please check on this etag field.

[imk13]

+1

[colgreen]

We are seeing this too, and have confirmed that the eTag property is the latest eTag using cosmos DB explorer. Seems like a bug in cosmos DB right now.

[manisrivastav]

@kushagraThapar we are having similar situation where, we are trying to replace the doc, we have retry logic by reading the latest but intermittently failing for some requests. Unpredictable and not reproducible in local.

JS SDK - @azure/cosmos": "3.6.2",

[v-dhampton]

This may be due to the Default Consistency level selected for your server which is likely Session. If a document read occurs within a couple milliseconds of document replacement from a different context, you may not receive the latest data.

You can read more about this here with more information here. The only guaranteed latest data consistency model is Strong but you probably don't want to switch.

However, this may not actually be your problem and you may have a real concurrency issue.