#cache #redis #memcache #twemproxy

bin+lib aster-proxy

A light, fast and powerful cache proxy written in Rust

17 stable releases

1.3.4 Jun 2, 2021
1.3.3 Jan 19, 2021
1.3.2 Jun 5, 2020
1.3.1 Mar 27, 2020
0.1.3 Nov 30, 2018

#395 in Caching

MIT license

255KB
7K SLoC

aster

Aster LOC

Aster is a lightweight, fast but powerful cache proxy written in rust.

It supports memcache/redis singleton/redis cluster protocol all in one. Aster can proxy with two models:

  1. proxy mode: the same as twemproxy.
  2. cluster mode: proxy for redis cluster. You can use non-cluster redis client access the redis cluster.(Inspired with Corvus)

Usage

cargo build --all --release && RUST_LOG=libaster=info RUST_BACKTRACE=1 ./target/release/aster default.toml

Configuration

[[clusters]]
# name of the cluster. Each cluster means one front-end port.

name="test-redis-cluster"

# listen_addr means the cluster font end serve address.

listen_addr="0.0.0.0:9001"

# cache_type only support memcache|redis|redis_cluster

cache_type="redis_cluster"

# servers means cache backend. support two format:
# for cache_type is memcache or redis, you can set it as:
#
#   servers = [
#       "127.0.0.1:7001:10 redis-1",
#       "127.0.0.1:7002:10 redis-2",
#       "127.0.0.1:7003:10 redis-3"]
#
# as you can see, the format is consisted with:
#
#       "${addr}:hash_weight ${node_alias}"
#
# And, for redis_cluster you can set the item as:
#
# servers = ["127.0.0.1:7000", "127.0.0.1:7001"]
#
# which means the seed nodes to connect to redis cluster.

servers = ["127.0.0.1:7000", "127.0.0.1:7001"]

# Work thread number, it's suggested as the number of your cpu(hyper-thread) number.

thread = 1

# ReadTimeout is the socket read timeout which effects all in the socket in millisecond

read_timeout = 2000

# WriteTimeout is the socket write timeout which effects all in the socket in millisecond

write_timeout = 2000

############################# Cluster Mode Special #######################################################
# fetch means fetch interval for backend cluster to keep cluster info become newer.
# default 10 * 60 seconds

fetch = 600


# read_from_slave is the feature make slave balanced readed by client and ignore side effects.
read_from_slave = true

############################# Proxy Mode Special #######################################################
# ping_fail_limit means when ping fail reach the limit number, the node will be ejected from the cluster
# until the ping is ok in future.
# if ping_fali_limit == 0, means that close the ping eject feature.

ping_fail_limit=3

# ping_interval means the interval of each ping was send into backend node in millisecond.

ping_interval=10000

changelog

see CHANGELOG.md

Dependencies

~44MB
~785K SLoC