Tyrant files

Tyrant provides you with three files for your application.

deployment-params.json

This file contains parameters that are used to control the deployment of your application. It should contain a valid JSON object.

Property name	Default value	Description
alarms	nil	An array of CloudWatch alarm definitions (a description is provided below).
autoScalingGroupDeletionAttempts	100	The number of times to check when determining whether an auto scaling group has been deleted.
defaultCooldown	10	The number of seconds after a scaling activity before another can start.
desiredCapacity	1	The desired number of instances to run in your Auto Scaling group.
ebsOptimized	false	Whether the instances launched should be EBS optimized.
healthCheckGracePeriod	600	The number of seconds after an instance launches during which it isn't considered for health checking by the Auto Scaling group.
healthCheckType	EC2	The type of healthcheck to use for the Auto Scaling group. If you use an EC2 healthcheck your instances will be kept in the Auto Scaling group if its status checks come back green. If those checks fail the instance will be terminated and another launched in its place. Using an ELB healthcheck will terminate the instance if it becomes unhealthy based on the healthcheck defined on the Load Balancer.
healthy	nil	An object describing the parameters to use with Healthy (a description is provided below). If this object is present we will assume that you want to use the Healthy service to check the healthy of the instances, which is handy if you don't have a load balancer but still want instances to go out of rotation in ZooKeeper.
instanceHealthyAttempts	50	The number of times to check when determining whether the launched instances are healthy.
instanceType	t1.micro	The Instance Type to launch. A complete listing of Instance Types can be found in Amazon's documentation.
loadBalancerHealthyAttempts	50	The number of times to check when determining whether the launched instances are healthy according to any load balancers they have been added to.
max	1	The minimum number of instances to run in your Auto Scaling group.
min	1	The maximum number of instances to run in your Auto Scaling group.
pauseAfterDeregisterOldInstances	false	Whether the deployment should pause after any old instances have been removed from the load balancer.
pauseAfterInstancesHealthy	false	Whether the deployment should pause after any instances are considered healthy.
pauseAfterLoadBalancersHealthy	false	Whether the deployment should pause after any load balancers are considered healthy.
policies	nil	An array of scaling policy definitions (a description is provided below).
scaleDownAfterDeployment	false	Whether, after deployment, your application should be scaled down to zero instances. Useful if you plan on adding scheduledActions.
scheduledActions	nil	A map of scheduled action name to the description of the scheduled action itself (a description is provided below).
selectedLoadBalancers	nil	An array containing the names of the Elastic Load Balancers you want to add your instances to. These Load Balancers will need to be running in the VPC.
selectedSecurityGroups	nil	An array containing the names (or IDs) of Security Groups to use for your instances. These need to have been created in the VPC.
selectedZones	nil	An array containing the suffix of the availability zones you want to run your instances in (e.g. ["a" "b"]). These will be appended to the name of the region. During deployment, if not provided, Exploud will use all zones available for the chosen subnetPurpose.
skipInstanceHealthcheck	false	Whether, during deployment, instance healthchecks should be skipped.
subnetPurpose	internal	A string representing the purpose of the subnet(s) the application should be deployed to.
subscriptions	nil	An array of strings listing the Sensu subscriptions which the instance should have.
terminationPolicy	Default	The policy for determining which instance should be terminated when scaling the Auto Scaling group down. Other values are ClosestToNextInstanceHour, NewestInstance, OldestInstance and OldestLaunchConfiguration. See Amazon's documentation for more details.
volumes	nil	An object which allows control over the volumes which will be added to each instance (a description is provided below).

Healthy

Additional descriptions are provided in the Healthy README.

Property name	Description
path	The relative path (from the root resource) which should be used for health checking.
port	The port Healthy should use to find the healthcheck resource. If provided, this will override any value provided in the application-properties.json service.port field.
scheme	The scheme to use for healthchecking (http or https).
timeout	An integer time (in seconds) to wait before deciding that something is unhealthy.

CloudWatch alarms

More detailed descriptions of these properties can be found in the PutMetricAlarmRequest documentation.

A note on alarm actions: these can either be the ARN of an SNS topic or a map which references a scaling policy (for example {"policy":"some-policy-name"}. The policy name referenced must be present in the policies array at the top level of the deployment-params. During deployment the policy will be created and its ARN will be switched into the array in place of the map.

Property name	Description
alarmActions	An array of actions to take when the alarm transitions into an ALARM state.
alarmDescription	A description of the alarm.
alarmName	The name of the alarm (must be unique for an auto-scaling group). Will have the name of the auto-scaling group prepended to it.
comparisonOperator	One of the comparison operators which determine how to compare the metric value to the threshold.
evaluationPeriods	The number of periods (as specified below) over which the metric should be evaluated.
insufficientDataActions	An array of actions to take when the alarm transitions into an INSUFFICIENT_DATA state.
metricName	The name of the metric to base this alarm on.
namespace	The namespace in which to find the metric.
okActions	An array of actions to take when the alarm transitions into an OK state.
period	The period (in seconds) over which the statistic is applied.
statistic	The statistic to apply to the metric value.
threshold	The threshold at which to take action (based on the comparison operator).
unit	The unit for alarm's associated metric.

Scaling policies

More detailed descriptions of these properties can be found in the PutScalingPolicyRequest documentation.

Property name	Description
adjustmentType	The type of adjustment this policy will be making.
cooldown	The number of seconds to allow between scaling activities.
estimatedInstanceWarmup	The number of seconds until a newly-launched instance can contribute to the metrics.
metricAggregationType	How to aggregate the metrics (not allowed with SimpleScaling policy type).
minAdjustmentMagnitude	The minimum number of instances to scale by (only used with PercentChangeInCapacity
policyName	The name of the policy, must be unique to each auto scaling group.
policyType	Either SimpleScaling (the default if not present) or StepScaling).
scalingAdjustment	The amount to scale by (only used with SimpleScaling policy type).
adjustmentType).
stepAdjustments	An array of step adjustment maps (described below) which govern how to scale (required with StepScaling policy type).

Step adjustments

These are probably described in much better detail in the AWS documentation.

Property name	Description
metricIntervalLowerBound	The lower bound between the alarm threshold and the current metric value for this step (a null value indicates negative infinity).
metricIntervalUpperBound	The upper bound between the alarm threshold and the current metric value for this step (a null value indicates positive infinity).
scalingAdjustment	How many instances to add.

Scheduled actions

Property name	Description
cron	A cron schedule indicating when to perform the scaling action.
desiredCapacity	Same as desiredCapacity for top-level deployment params.
max	Same as max for top-level deployment params.
min	Same as min for top-level deployment params.

Volumes

Property name	Description
root	An EBS description for the root disk (will be mounted as /dev/xvda on HVM instances and /dev/sda1 on paravirtual instances).
instanceStores	The number of ephemeral instance-storage devices to add (will default to the number available to the instance type chosen.
blockDevices	An array containing EBS descriptions for an EBS devices to be added.

EBS description

Property name	Description
deleteOnTermination	Whether the device should be deleted when the instance is terminated.
iops	The number of IOPS to be assigned to the device (used with io1 type).
size	The size (in GB) of the device.
snapshotId	The ID of a snapshot to base this device on.
type	The type of device to use (gp2 for general purpose SSD, io1 for provisioned IOPS SSD, standard for magnetic disk).

Sample

{
  "healthCheckType" : "ELB",
  "scaleDownAfterDeployment" : true,
  "scheduledActions" : {
    "scale-down" : {
      "cron" : "30 12 * * *",
      "desiredCapacity" : 0,
      "max" : 0,
      "min" : 0
    },
    "scale-up" : {
      "cron" : "30 4 * * *",
      "desiredCapacity" : 1,
      "max" : 1,
      "min" : 1
    }
  },
  "selectedLoadBalancers" : [
    "recommendations"
  ],
  "selectedZones" : [
    "a"
  ],
  "selectedSecurityGroups" : [
    "Brislabs-8080",
    "Brislabs-SSH"
  ],
  "subscriptions" : [
    "healthcheck-8080",
    "java"
  ],
  "volumes" : {
    "root" : {
      "size" : 20
    },
    "instanceStores" : 0,
    "blockDevices" : [
      {
        "size" : 500,
        "type" : "gp2"
      }
    ]
  }
}

application-properties.json

This file contains the properties required for your application to run. It should contain a valid JSON object. There is no special place for the old environment properties in the new deployment tools, they should be specified like any other.

Sample

{
  "logging.level" : "WARN",
  "mongo.host" : "localhost"
}

launch-data.json

This file allows you to specify commands which will be executed (as root) after your properties have been written to /etc/{application}.properties. Tyrant will pre-populate this file with a single command which will create a Beaver configuration file which will ship your application logs via Logstash.

Sample

[
  "first command",
  "second command"
]