read-kinesis

read-kinesis

Read records from the shards of an AWS Kinesis stream and dump them to standard output.

Usage

> read-kinesis dump <stream-name> [options]

> # Read from all shards in the specified stream (based on list-shards)
> read-kinesis dump $STREAM_NAME --all

> # Read from specified shards in the stream
> read-kinesis dump $STREAM_NAME --shard shardId-000000000000 --shard shardId-000000000001

Long Wait

If you have a very sparse stream (relatively few records on it), it might take a long time to start getting any records, or even just to determine there are no records. This has to do with the way Kinesis timeslices the stream and iterates over timeslices. If you don't have any checkpoints, or your check point is from a long time ago, the reader may have to read over a bunch of empty timeslices before reaching any records or the end of the stream. You can add the --verbose flag to get some status output to STDERR to see what's going on.

Output

For each record read from Kinesis, the record is presented with all of it's fields and some additional fields beginning with a _ prefix. These fields are as follows:

Record field	Description
_shardId	The shard it came from (see note 1, below)
_streamName	The name of the stream
_data	The data payload of the record (see note 2, below)

Notes:

1. If the --shard option is used, the value of the __shardId field is the value given in the option. Kinesis supports some interpretation of the shardId, so multiple values may alias to the same shard. E.g., --shard 0 and --shard 000 both refer to the shard whose actual shard ID is "shardId-000000000000".

2. The Data field that comes from Kinesis is not shown as it's not particularly useful, being replaced instead with the _data field whose value depends on the --data-format as described in the "Data Format" section.

Data Format

Kinesis record payloads are binary data. By default, this tool assumes the binary data is utf8 encoded text, and attempts to decode it as such. You can change the behavior using the --data-format option (-d for short). The following options are available:

--data-format arguments	Description
bin	Presents the data as a Buffer of octets
hex	Presents the data as a hex string
base64	Presents the data as a base64 string
utf8	Parses the data as utf8 and presents the text
json	Parses the data as utf8 encoded JSON and presents it as the JSON value (not a string)

Checkpoints

A checkpoint describes where the reader leftoff in a particular shard (in a particular stream). You can add the --checkpoint option to specify that checkpoints should be used. When given, checkpoints are loaded from the checkpoint file, if it exists (the default checkpoint file is .checkpoints in the current directory, you can use the --checkpoint-file option to specify an alternate). If no checkpoints are found for a shard, it starts at the trim horizon (i.e., the oldest record currently in the shard).

After reading completes successfully, if the --checkpoint option is given, then updated checkpoints are appended to the checkpoint file, indicating the furthest that the command read in each shard so that subsequent commands can pick up where it left off.

The checkpoint file is appended to by default, so it could potentially get really big. Use the --trim-checkpoints option to overwrite the entire file, instead.

Credentials / Authorization

You'll need to have AWS credentials authorized for reading from the specified Kinesis stream. There are various ways to provide these; see "Setting Credentials in Node.js" in the AWS documentation for details. Here's a quick summary of your options (in order of precedence):

1. Define the AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY environment variables (and optionally AWS_SESSION_TOKEN)
2. Your credentials file (usually ~/.aws/credentials/)

If you specify the --profile option to the command, it will use the specified profile from your credentials file.

You can also use the --assume option to specify the ARN of an IAM role to assume. Specify the option multiple times to specify a chain of roles that should be assumed. For instance, if you're authorized to assume role A but not role B, but role A is authorized to assume role B, you can use role B with --assume A --assume B.

CLI Options

Option	Description
-r REGION --region REGION	The AWS region of the stream. required
-s SHARD-ID --shard SHARD-ID	Specify the name of the shard to read from. Give this option multipe times to read from multiple shards.
-a --all	Read from all shards in the given stream
-t TIME --timestamp TIME	Specify a time to start reading from. You can use an ISO-8601 date (e.g., "2020-01-30" or "2020-01-30T12:34:56-0400"), or a "human" duration as parsable by parse-human-relative-time (e.g, "1 hour ago", "last monday"). Note that this will ignore any existing checkpoints, but wil still update checkpoints at the end. This means if you specified time is later than your previous checkpoints, you'll end up missing the records in between the last checkpoint and the timestamp.
-c --checkpoint	Read and use initial checkpoints from file, if present. Write checkpoints to file if everything completes successful. Use the --checkpoint-file to specify the file to use, the default is .checkpoints.
--profile	Use the specified profile from your shared credentials file (typically ~/.aws/credentials) for AWS credentials.
--assume	Assume the AWS role specified by this ARN for reading from Kinesis. You can specify this option multiple times to specify a chain of roles that will be assumed
--json	Output the records in pretty-printed JSON. The default is to use console.log to format the output
--jsonl --json-lines	Output the records in JSON lines, one line per record
-d FORMAT --data-format FORMAT	Specifies how to handle the data payload of kinesis records. See "Data Format" section above
--checkpoint-file FILE	Specify the path to the checkpoint file to use. Only relevant if the --checkpoint option is given. (default: .checkpoints)
--trim-checkpoints	Overwrite the contents of the checkpoint-file instead of appending to it.
--stop-after TIMESTAMP	Optionally specify a timestamp to stop reading after. Applies individually to each shard read. Reader will read and output an entire batch of records at a time as usual, but will terminate once it gets a batch whose last record is greater than the specified timestamp. Note that this is based on the records "ApproximateArrivalTimestamp" value.