Kinesis GetRecords truncates ApproximateArrivalTimestamp field

[dacut]

Confirm by changing [ ] to [x] below to ensure that it's a bug:

• I've gone though Developer Guide and API reference
• I've checked AWS Forums and StackOverflow for answers
• I've searched for previous similar issues and didn't find any solution

Describe the bug
When calling kinesis.GetRecords() for a shard, the ApproximateArrivalTimestamp field which is stored with millisecond precision per the Kinesis documentation is truncated to have zero fractional seconds. This doesn't happen for the same records with the Boto3 (Python) SDK.

Printing the ApproximateArrivalTimestamp field (to nanosecond resolution, e.g. time.Format("15:04:05.999999999")) for the Go vs Boto3 SDK yields:

Go                  Boto3
07:59:13.000000000  07:59:13.913000000
07:59:19.000000000  07:59:19.112000000
07:59:33.000000000  07:59:33.929000000
07:59:39.000000000  07:59:39.133000000
07:59:51.000000000  07:59:51.735000000
07:59:53.000000000  07:59:53.932000000
07:59:59.000000000  07:59:59.153000000
08:00:13.000000000  08:00:13.965000000
08:00:19.000000000  08:00:19.171000000
08:00:33.000000000  08:00:33.985000000
08:00:39.000000000  08:00:39.192000000
08:00:51.000000000  08:00:51.680000000
08:00:53.000000000  08:00:53.987000000
08:00:59.000000000  08:00:59.210000000

And, yes, we do need sub-second resolution for our application logs.

Version of AWS SDK for Go?
Example: v1.32.13

• get SDK version by printing the output of aws.SDKVersion in your code after importing "github.com/aws/aws-sdk-go/aws"

Version of Go (go version)?
go version go1.14.4 darwin/amd64

To Reproduce (observed behavior)
Actual code sample:
Obtaining Kinesis records example: https://github.com/dacut/kinesis-log-watcher/blob/master/main.go#L217
And printing them: https://github.com/dacut/kinesis-log-watcher/blob/master/main.go#L247

Simplified example:

package main

import (
	"fmt"
	"time"

	"github.com/aws/aws-sdk-go/aws"
	"github.com/aws/aws-sdk-go/aws/session"
	"github.com/aws/aws-sdk-go/service/kinesis"
)

func main() {
	awsSession := session.Must(session.NewSession())
	kinesisClient := kinesis.New(awsSession)
	gsiInput := kinesis.GetShardIteratorInput{
		StreamName: aws.String("WebHost-development-Logs"),
		ShardId: aws.String("shardId-000000000000"),
		ShardIteratorType: aws.String("TRIM_HORIZON")
	}
	gsiOutput, _ := kinesisClient.GetShardIterator(&gsiInput)

	grInput := kinesis.GetRecordsInput{ShardIterator: gsiOutput.ShardIterator}
	for {
		grOutput, _ := kinesisClient.GetRecords(&grInput)
		for _, record := range grOutput.Records {
			ts := aws.TimeValue(record.ApproximateArrivalTimestamp)
			fmt.Printf("Timestamp=%s Nanoseconds=%d\n", ts.Format(time.RFC3339Nano), ts.Nanosecond())
		}

		if aws.Int64Value(grOutput.MillisBehindLatest) == 0 {
			break
		}

		grInput.ShardIterator = grOutput.NextShardIterator
	}
}

Expected behavior
The nanoseconds field of timestamps should be non-zero (though with only millisecond resolution).

Additional context
None.

[diehlaws]

Hi @dacut, thanks for bringing this to our attention and I do apologize for the delay in response on our end. I'm able to reproduce the described behavior and have brought this up with the rest of the team, we'll be investigating this behavior shortly. Once we have additional information on the matter we'll update the issue accordingly.