scale.io

loadbalancing and autoscaling module for websocket based EC2 cluster without ELB.

architecture

• no ELB

• slack based monitoring

• zero downtime deploy
• multi-AZ based
• auto fail over

how to use

scale.js

require('scale.io');

set config yml file

set config/default.yaml or config/env.yaml your main app.

Example is this.

execute

node --harmony scale.js

config details

conf for redis / aws / slack

enter your

• redis host & port
• access key id, secret access key and region
• slack info

conf for loadbalancing / autoscaling / monitoring

enter some info (details are below).

conf for loadbalancing

Correspond to A and B in above structure chart.

|key|desc| |---|---| |timeoutThresholdForRedHost|Getting no response within this value(ms), host is regard as RedHost(=DyingHost).| |connNumForRedHost|RedHost is expressed by this value at loadbalancing log.| |ec2HostTagKey|EC2 tag key| |ec2HostTagValueForLobby|EC2 tag value for lobby server.| |ec2HostTagValueForGame|EC2 tag value for game server.| |intervalForAutoscalingSec|cycle time of autoscale watching process| |intervalForHostInfoUpdateSec|cycle time of loadbalancing process| |redisKeyForLobbyLoadbalancing|redis key name for lobby loadbalance.| |redisKeyForGameLoadbalancing|redis key name for game loadbalance.|

conf for autoscaling

Correspond to C in above structure chart.

|key|desc| |---|---| |cfnStackName|stack name of cloudformation.| |serverTypeNum|server type number. if you use lobby and game, enter 2.| |scaleOutThreshold|average connection number of instances for scale out.| |scaleInThreshold|average connection number of instances for scale in.| |cfnUpdateCmd|update cloudformation command.| |minimumGreenHostsNum|falling below this value, scale out starts.| |forceScaleOutStartAt|scale out forced to start after this time.| |forceScaleOutEndAt|scale out forced to start before this time.| |scaleInAvailableStartAt|scale in can be started after this time.| |scaleInAvailableEndAt|scale in can be started after this time.| |notifyRoomNumMessage|notifying message by slack.| |scaleInMinimumIntervalSec|scale in can be started countinuously after this interval.| |scaleOutMinimumIntervalSec|scale out can be started countinuously after this interval.| |specifiedTimeGreenHostsNum|instance number will be change to this value between specified time.| |restartCmd|restart autoscale process commond.| |autoRestartCpuUsageThresholdForLB|if cpu usage exceed this value, process restart.| |autoRestartMonitorIntervalSec|monitoring span for auto restart.|

conf for monitor

Correspond to D in above structure chart.

|key|desc| |---|---| |monitoredHost|monitored target.| |monitoredProcess|like 'game', 'lobby'| |monitoringIntervalSec|monitoring span| |decisionIntervalSec|monitored process does not reply throughout this span, it is regard as died.|