GoFlow

This application is a NetFlow/IPFIX/sFlow collector in Go.

It gathers network information (IP, interfaces, routers) from different flow protocols, serializes it in a protobuf format and sends the messages to Kafka using Sarama's library.

Why

The diversity of devices and the amount of network samples at Cloudflare required its own pipeline. We focused on building tools that could be easily monitored and maintained. The main goal is to have full visibility of a network while allowing other teams to develop on it.

Modularity

In order to enable load-balancing and optimizations, the GoFlow library has a decoder which converts the payload of a flow packet into a Go structure.

The producer functions (one per protocol) then converts those structures into a protobuf (pb/flow.pb) which contains the fields a network engineer is interested in. The flow packets usually contains multiples samples This acts as an abstraction of a sample.

The transport provides different way of processing the protobuf. Either sending it via Kafka or print it on the console.

Finally, utils provide functions that are directly used by the CLI utils. GoFlow is a wrapper of all the functions and chains thems into producing bytes into Kafka. There is also one CLI tool per protocol.

You can build your own collector using this base and replace parts:

• Use different transport (eg: RabbitMQ instead of Kafka)
• Convert to another format (eg: Cap'n Proto, Avro, instead of protobuf)
• Decode different samples (eg: not only IP networks, add MPLS)
• Different metrics system (eg: use expvar instead of Prometheus)

Protocol difference

The sampling protocols can be very different:

sFlow is a stateless protocol which sends the full header of a packet with router information (interfaces, destination AS) while NetFlow/IPFIX rely on templates that contain fields (eg: source IPv6).

The sampling rate in NetFlow/IPFIX is provided by Option Data Sets. This is why it can take a few minutes for the packets to be decoded until all the templates are received (Option Template and Data Template).

Both of these protocols bundle multiple samples (Data Set in NetFlow/IPFIX and Flow Sample in sFlow) in one packet.

The advantages of using an abstract network flow format, such as protobuf, is it enables summing over the protocols (eg: per ASN or per port, rather than per (ASN, router) and (port, router)).

Features

Collection:

• NetFlow v5
• IPFIX/NetFlow v9
  • Handles sampling rate provided by the Option Data Set
• sFlow v5: RAW, IPv4, IPv6, Ethernet samples, Gateway data, router data, switch data

Production:

• Convert to protobuf
• Sends to Kafka producer
• Prints to the console

Monitoring:

• Prometheus metrics
• Time to decode
• Samples rates
• Payload information
• NetFlow Templates

Run

Download the latest release and just run the following command:

./goflow -h

Enable or disable a protocol using -nf=false or -sflow=false. Define the port and addresses of the protocols using -nf.addr, -nf.port for NetFlow and -sflow.addr, -slow.port for sFlow.

Set the brokers or the Kafka brokers SRV record using: -kafka.brokers 127.0.0.1:9092,[::1]:9092 or -kafka.srv. Disable Kafka sending -kafka=false. You can hash the protobuf by key when you send it to Kafka.

You can collect NetFlow/IPFIX, NetFlow v5 and sFlow using the same collector or use the single-protocol collectors.

You can define the number of workers per protocol using -workers .

Docker

We also provide a all-in-one Docker container. To run it in debug mode without sending into Kafka:

$ sudo docker run --net=host -ti cloudflare/goflow:latest -kafka=false

Environment

To get an example of pipeline, check out flow-pipeline

How is it used at Cloudflare

The samples flowing into Kafka are processed and special fields are inserted using other databases:

• User plan
• Country
• ASN and BGP information

The extended protobuf has the same base of the one in this repo. The compatibility with other software is preserved when adding new fields (thus the fields will be lost if re-serialized).

Once the updated flows are back into Kafka, they are consumed by database inserters (Clickhouse, Amazon Redshift, Google BigTable...) to allow for static analysis. Other teams access the network data just like any other log (SQL query).

Output format

If you want to develop applications, build pb/flow.proto into the language you want:

Example in Go:

PROTOCPATH=$HOME/go/bin/ make proto

Example in Java:

export SRC_DIR="path/to/goflow-pb"
export DST_DIR="path/to/java/app/src/main/java"
protoc -I=$SRC_DIR --java_out=$DST_DIR $SRC_DIR/flow.proto

The fields are listed in the following table.

You can find information on how they are populated from the original source:

• For sFlow
• For NetFlow v5
• For NetFlow v9
• For IPFIX

If you are implementing flow processors to add more data to the protobuf, we suggest you use field IDs ≥ 1000.

Implementation notes

The pipeline at Cloudflare is connecting collectors with flow processors that will add more information: with IP address, add country, ASN, etc.

For aggregation, we are using Materialized tables in Clickhouse. Dictionaries help correlating flows with country and ASNs. A few collectors can treat hundred of thousands of samples.

We also experimented successfully flow aggregation with Flink using a Keyed Session Window: this sums the Bytes x SamplingRate and Packets x SamplingRate received during a 5 minutes window while allowing 2 more minutes in the case where some flows were delayed before closing the session.

The BGP information provided by routers can be unreliable (if the router does not have a BGP full-table or it is a static route). You can use Maxmind prefix to ASN in order to solve this issue.

License

Licensed under the BSD 3 License.