Kafka#

The Kafka ingester designed to act as a consumer for Apache Kafka so that data Gravwell can attach to a Kafka cluster and consume data. Kafka can act as a high availability data broker to Gravwell. Kafka can take on some of the roles provided by the Gravwell Federator, or ease the burden of integrating Gravwell into an existing data flow. If your data is already flowing to Kafka, integrating Gravwell is just an apt-get away.

The Gravwell Kafka ingester is best suited as a co-located ingest point for a single indexer. If you are operating a Kafka cluster and a Gravwell cluster, it is best not to duplicate the load balancing characteristics of Kafka at the Gravwell ingest layer. Install the Kafka ingester on the same machine as the Gravwell indexer and use the Unix named pipe connection. Each indexer should be configured with its own Kafka ingester, this way the Kafka cluster can manage load balancing.

Most Kafka configurations enforce a data durability guarantee, which means data is stored in non-volatile storage when consumers are not available to consume it. As a result we do not recommend that the Gravwell ingest cache be enabled on Kafka ingester, instead let Kafka provide the data durability.

Installation#

To install the Debian package, make sure the Gravwell Debian repository is configured as described in the quickstart. Then run the following command as root:

apt update && apt install gravwell-kafka

To install the Redhat package, make sure the Gravwell Redhat repository is configured as described in the quickstart. Then run the following command as root:

yum install gravwell-kafka

To install via the standalone shell installer, download the installer from the downloads page, then run the following command as root, replacing X.X.X with the appropriate version:

bash gravwell_kafka_installer_X.X.X.sh

You may be prompted for additional configuration during the installation.

The Docker image is available on Dockerhub.

Basic Configuration#

The Kafka ingester uses the unified global configuration block described in the ingester section. Like most other Gravwell ingesters, the Kafka Ingester supports multiple upstream indexers, TLS, cleartext, and named pipe connections, a local cache, and local logging.

Consumer Examples#

[Consumer "default"]
	Leader="127.0.0.1"
	Default-Tag=default   #send bad tag names to default tag
	Tags=*                #allow all tags
	Topic=default
	Tag-Header=TAG        #look for the tag in the Kafka TAG header
	Source-Header=SRC     #look for the source in the Kafka SRC header

[Consumer "test"]
	Leader="127.0.0.1:9092"
	Tag-Name=test
	Topic=test
	Consumer-Group=mygroup
	Synchronous=true
	Key-As-Source=true #A custom feeder is putting its source IP in the message key value
	Header-As-Source="TS" #look for a header key named TS and treat that as a source
	Source-As-Text=true #the source value is going to come in as a text representation
	Batch-Size=256 #get up to 256 messages before consuming and pushing
	Rebalance-Strategy=roundrobin

Installation#

The Kafka ingester is available in the Gravwell Debian repository as a Debian package as well as a shell installer on our Downloads page. Installation via the repository is performed using apt:

apt-get install gravwell-kafka

The shell installer provides support for any non-Debian system that uses systemd, including Arch, Redhat, Gentoo, and Fedora.

root@gravserver ~ # bash gravwell_kafka_installer.sh

Configuration#

The Gravwell Kafka ingester can subscribe to multiple topics and even multiple Kafka clusters. Each consumer defines a consumer block with a few key configuration values.

Parameter	Type	Descriptions	Required
Tag-Name	string	The Gravwell tag that data should be sent to.	YES
Leader	host:port	The Kafka cluster leader/broker. This should be an IP or hostname, if no port is specified the default port of 9092 is appended	YES
Topic	string	The Kafka topic this consumer will read from	YES
Consumer-Group	string	The Kafka consumer group this ingester is a member of	NO - default is `gravwell`
Source-Override	IPv4 or IPv6	An IP address to use as the SRC for all entries	NO
Rebalance-Strategy	string	The re-balancing strategy to use when reading from Kafka	NO - default is `roundrobin`. `sticky`, and `range` are also options
Key-As-Source	boolean	Gravwell producers will often put the data source address in a message key, if set the ingester will attempt to interpret the message key as a Source address. If the key structure is not correct the ingester will apply the override (if set) or the default source.	NO - default is false
Synchronous	boolean	The ingester will perform a sync on the ingest connection every time a Kafka batch is written.	NO - default is false
Batch-Size	integer	The number of entries to read from Kafka before forcing a write to the ingest connection	NO - default is 512
Auth-Type	string	Enable SASL authentiation and specify mechanism
Username	string	Specify username for SASL authentication
Password	string	Specify password for SASL authentication

Warning

Setting any consumer as synchronous causes that consumer to continually Sync the ingest pipeline. It will have significant performance implications for ALL consumers.

Note

Setting a large Batch-Size when using Synchronous=true can help with performance under heavy load.

Authentication#

The Kafka ingester supports the following SASL authentication mechanisms: PLAIN, SCRAMSHA256, and SCRAMSHA512. To enable authentication set the Auth-Type configuration parameter to the desired authentication type and provide a Username and Password. Each consumer must set its own authentication.

The following is a valid configuration snippet from a listener named tester using plaintext authentication:

[Consumer "tester"]
	Leader="127.0.0.1"
	Default-Tag=default   #send bad tag names to default tag
	Tags=*                #allow all tags
	Topic=default
	Auth-Type=plain
	Username=TheDude
	Password="a super secret password"

Valid Auth-Type options are plain, scramsha256, and scramsha512.

Example Configuration#

Here is an example configuration that is subscribing to two different topics using two different consumer groups.

[Global]
Ingest-Secret = IngestSecrets
Connection-Timeout = 0
Pipe-Backend-target=/opt/gravwell/comms/pipe
Log-Level=INFO
Log-File=/opt/gravwell/log/kafka.log

[Consumer "default"]
	Leader="tasks.kafka.internal"
	Tag-Name=default
	Topic=default
	Consumer-Group=gravwell1
	Key-As-Source=true
	Batch-Size=256


[Consumer "test"]
	Leader="tasks.testcluster.internal:9092"
	Tag-Name=test
	Topic=test
	Consumer-Group=testgroup
	Source-Override="192.168.1.1"
	Rebalance-Strategy=range
	Batch-Size=4096