tpc

tpc roughly stands for Twemproxy Configurator

Listens to a cluster of redis sentinel servers via pubsub. Rewrites twemproxy config file and executes twemproxy reload command when sentinel notifies of redis master failover.

Install

Download the latest release, extract it, and put it somewhere on your PATH.

or

$ go get github.com/atongen/tpc

or

$ mkdir -p $GOPATH/src/github.com/atongen
$ cd $GOPATH/src/github.com/atongen
$ git clone git@github.com:atongen/tpc.git
$ cd tpc
$ go install
$ rehash

Testing

$ cd $GOPATH/src/github.com/atongen/tpc
$ go test -cover

Releases

$ mkdir -p $GOPATH/src/github.com/atongen
$ cd $GOPATH/src/github.com/atongen
$ git clone git@github.com:atongen/tpc.git
$ cd tpc
$ make release

Command-Line Options

λ tpc -h
Usage of tpc:
  -auto_eject_hosts
        Twemproxy: A boolean value that controls if server should be ejected temporarily when it fails consecutively server_failure_limit times.
  -backlog int
        Twemproxy: TCP backlog argument (default 1024)
  -channel string
        Slack: channel for notifications (default "#incidents")
  -client_connections int
        Twemproxy: The maximum number of connections allowed from redis clients (default 4096)
  -cmd string
        Process: Command to execute after master failover
  -distribution string
        Twemproxy: Key distribution (default "ketama")
  -has_tag string
        Twemproxy: A two character string that specifies the part of the key used for hashing. Eg '{}' or '$$'
  -hash string
        Twemproxy: Hash algorithm (default "fnv1a_64")
  -icon_emoji string
        Slack: icon emoji for notifications
  -ip string
        Twemproxy: Ip address (default "0.0.0.0")
  -listen_address string
        Prometheus: Listen address (default ":9298")
  -log string
        Process: Path to log file, will write to STDOUT if empty
  -master_pattern string
        Process: If provided, will filter master names from sentinel based on pattern
  -name string
        Twemproxy: Name of redis pool (default "redis")
  -out string
        Process: File to write configuration, will write to STDOUT if empty
  -backup string
        Process: Directory to write backup files, will write to STDOUT if empty
  -port int
        Twemproxy: Port (default 9000)
  -preconnect
        Twemproxy: A boolean value that controls if we should preconnect to all the servers in this pool on process start (default true)
  -redis_auth string
        Twemproxy: authenticate to the redis server on connect
  -redis_db int
        Twemproxy: The DB number to use on the redis pool servers. Twemproxy will always present itself to clients as DB 0
  -sentinels string
        Process: CSV of host:port to redis sentinels
  -server_connections int
        Twemproxy: The maximum number of connections that can be open to each server (default 1)
  -server_failure_limit int
        Twemproxy: The number of consecutive failures on a server that would lead to it being temporarily ejected when auto_eject_host is set to true. (default -1)
  -server_retry_timeout int
        Twemproxy: The timeout value in msec to wait for before retrying on a temporarily ejected server, when auto_eject_host is set to true. (default -1)
  -telemetry_path string
        Prometheus: Telemetry path (default "/metrics")
  -timeout int
        Twemproxy: The timeout value in msec that we wait for to establish a connection to the server or receive a response from a server. (default -1)
  -token string
        Slack: API token used for notifications
  -username string
        Slack: username for notifications
  -v    Print version information and exit

Issues

twemproxy hot reload

There is a long-running twemproxy feature branch that is supposed to bring config hot-reload via unix signal:

• twitter/twemproxy#6
• twitter/twemproxy#321
• https://github.com/machinezone/twemproxy/tree/lwalkin/config-reload

Currently the solution is to pause the redis clients, then stop and restart the twemproxy process.