This Go code implements a client-server message-queue system like Kafka but simpler.
- The server stores Messages in topics.
- Messages are just byte sequences; of arbitrary-length.
- Clients can post messages to a topic using the Produce client library.
- Other Clients can subscribe to messages as they arrive using the Consumer client library.
- Messages are not removed from the server when they are consumed.
- Thus, individual consuming clients can consume the stream at their own rate.
- But Messages are removed from the server when they reach a configurable age; specified in anything from milliseconds to days.
Messages can either be stored to a file-system, or (volatile) memory.
Read the rationale for using the file-system for storage.
The underlying client/server communciations uses gRPC (unlike Kafka's custom tcp protocol). But the client libraries hide that away, and expose a few simple API methods you can call on proxy objects.
- It works
- It has decent unit test coverage
- The design, code and documentation is respectable, but has not been peer reviewed yet.
- It is still a work in progress
- It hasn't been benchmarked, tuned or stress-tested yet.
go get google.golang.org/grpc
go get github.com/peterhoward42/minikafka
cd $GOPATH/src/github.com/peterhoward42/minikafka
go install ./...
export MINIKAFKA_HOST=":9999"
export MINIKAFKA_RETENTIONTIME="10s"
export MINIKAFKA_ROOT_DIR="/tmp/minikafka"
mkfk-server
If you want to use the in-memory store instead of a file-system store:
export MINIKAFKA_ROOT_DIR=""
You can try out a simple command line wrapper to the client library:
mkfk-producer -host localhost:9999 -topic topic_foo
This sends each line of text you enter to the server as a Produce message. See below for the more realistic option of embedding the producer (and consumer) client in your own code.
mkfk-consumer -host localhost:9999 -topic topic_foo
This polls the server every 3 seconds and tells you what it got back. Remember though, that the messages only live on the server with these settings for 10 seconds.
The more realistic use-case is to incorporate a producer or consumer client library in your own app - as illustrated by the command line consumer wrapper code., or the producer wrapper code.
You can similarly wrap the server library in your own code, perhaps to obtain the configuration from something other than environment variables. See the server wrapper code.
The beauty of gRPC is that you can auto-generate client code in most languages using the protobuf file here, and the gRPC tools.