Monitor GameLift with Amazon CloudWatch

You can monitor GameLift using Amazon CloudWatch, an AWS service that collects raw data and processes it into readable, near real-time metrics. These statistics are kept for 15 months to provide a historical perspective on how your game server hosting with GameLift is performing. You can set alarms that watch for certain thresholds and send notifications or take actions when those thresholds are met. For more information, see the Amazon CloudWatch User Guide.

The following tables list the metrics and dimensions for GameLift. All metrics that are available in CloudWatch are also available in the GameLift console, which provides the data as a set of customizable graphs. To access CloudWatch metrics for your games, use the AWS Management Console, the AWS CLI, or the CloudWatch API.

If a metric does not have a location, it uses the home location.

Dimensions for GameLift metrics

GameLift supports filtering metrics by the following dimensions.

Dimension	Description
Location	Filter metrics for a fleet deployment location. If a metric does not have a location, it uses the home location.
FleetId	Filter metrics for a single fleet. This dimension can be used with all fleet metrics for instances, server processes, game sessions, and player sessions.
MetricGroup	Filter metrics for a collection of fleets. Add a fleet to a metric group by adding the metric group name to the fleet's attributes (see UpdateFleetAttributes()). This dimension can be used with all fleet metrics for instances, server processes, game sessions, and player sessions.
QueueName	Filter metrics for a single queue. This dimension is used with metrics for game session queues only.
MatchmakingConfigurationName	Filter metrics for a single matchmaking configuration. This dimension is used with metrics for matchmaking configurations.
MatchmakingConfigurationName-RuleName	Filter metrics for an intersect of a matchmaking configuration and matchmaking rule. This dimension is used with metrics for matchmaking rules only.
InstanceType	Filter metrics for an EC2 instance type designation, such as "c4.large". This dimension is used with metrics for spot instances.
OperatingSystem	Filter metrics for an instance's operating system This dimension is used with metrics for spot instances.
GameServerGroup	Filter FleetIQ metrics for a game server group.

GameLift metrics for fleets

The AWS/GameLift namespace includes the following metrics related to activity across a fleet or a group of fleets. Fleets are used with a managed GameLift solution. The GameLift service sends metrics to CloudWatch every minute.

Instances

Metric	Description
ActiveInstances	Instances with ACTIVE status, which means they are running active server processes. The count includes idle instances and those that are hosting one or more game sessions. This metric measures current total instance capacity. This metric can be used with automatic scaling. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum Dimensions: Location
DesiredInstances	Target number of active instances that GameLift is working to maintain in the fleet. With automatic scaling, this value is determined based on the scaling policies currently in force. Without automatic scaling, this value is set manually. This metric is not available when viewing data for fleet metric groups. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum Dimensions: Location
IdleInstances	Active instances that are currently hosting zero (0) game sessions. This metric measures capacity that is available but unused. This metric can be used with automatic scaling. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum Dimensions: Location
MaxInstances	Maximum number of instances that are allowed for the fleet. A fleet's instance maximum determines the capacity ceiling during manual or automatic scaling up. This metric is not available when viewing data for fleet metric groups. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum Dimensions: Location
MinInstances	Minimum number of instances allowed for the fleet. A fleet's instance minimum determines the capacity floor during manual or automatic scaling down. This metric is not available when viewing data for fleet metric groups. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum Dimensions: Location
PercentIdleInstances	Percentage of all active instances that are idle (calculated as IdleInstances / ActiveInstances). This metric can be used for automatic scaling. Units: Percent Relevant CloudWatch statistics: Average, Minimum, Maximum Dimensions: Location
RecycledInstances	Number of spot instances that have been recycled and replaced. GameLift recycles spot instances that are not currently hosting game sessions and have a high probability of interruption. Units: Count Relevant CloudWatch statistics: Sum, Average, Minimum, Maximum Dimensions: Location
InstanceInterruptions	Number of spot instances that have been interrupted. Units: Count Relevant CloudWatch statistics: Sum, Average, Minimum, Maximum Dimensions: Location

Server processes

Metric	Description
ActiveServerProcesses	Server processes with ACTIVE status, which means they are running and able to host game sessions. The count includes idle server processes and those that are hosting game sessions. This metric measures current total server process capacity. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum Dimensions: Location
HealthyServerProcesses	Active server processes that are reporting healthy. This metric is useful for tracking the overall health of the fleet's game servers. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum Dimensions: Location
PercentHealthyServerProcesses	Percentage of all active server processes that are reporting healthy (calculated as HealthyServerProcesses / ActiveServerProcesses). Units: Percent Relevant CloudWatch statistics: Average, Minimum, Maximum Dimensions: Location
ServerProcessAbnormalTerminations	Server processes that were shut down due to abnormal circumstances since the last report. This metric includes terminations that were initiated by the GameLift service. This occurs when a server process stops responding, consistently reports failed health checks, or does not terminate cleanly (by calling ProcessEnding()). Units: Count Relevant CloudWatch statistics: Sum, Average, Minimum, Maximum Dimensions: Location
ServerProcessActivations	Server processes that successfully transitioned from ACTIVATING to ACTIVE status since the last report. Server processes cannot host game sessions until they are active. Units: Count Relevant CloudWatch statistics: Sum, Average, Minimum, Maximum Dimensions: Location
ServerProcessTerminations	Server processes that were shut down since the last report. This includes all server processes that transitioned to TERMINATED status for any reason, including normal and abnormal process terminations. Units: Count Relevant CloudWatch statistics: Sum, Average, Minimum, Maximum Dimensions: Location

Game sessions

Metric	Description
ActivatingGameSessions	Game sessions with ACTIVATING status, which means they are in the process of starting up. Game sessions cannot host players until they are active. High numbers for a sustained period of time may indicate that game sessions are not transitioning from ACTIVATING to ACTIVE status. This metric can be used with automatic scaling. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum Dimensions: Location
ActiveGameSessions	Game sessions with ACTIVE status, which means they are able to host players, and are hosting zero or more players. This metric measures the total number of game sessions currently being hosted. This metric can be used with automatic scaling. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum Dimensions: Location
AvailableGameSessions	Active, healthy server processes that are not currently being used to host a game session and can start a new game session without a delay to spin up new server processes or instances. This metric can be used with automatic scaling. Note For fleets that limit concurrent game session activations, use the metric ConcurrentActivatableGameSessions. That metric more accurately represents the number of new game sessions that can start without any type of delay. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum Dimensions: Location
ConcurrentActivatableGameSessions	Active, healthy server processes that are not currently being used to host a game session and can immediately start a new game session. This metric differs from AvailableGameSessions in the following way: it does not count server processes that currently cannot activate a new game session because of limits on game session activations. (See the fleet RuntimeConfiguration optional setting MaxConcurrentGameSessionActivations). For fleets that don't limit game session activations, this metric is identical to AvailableGameSessions. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum Dimensions: Location
PercentAvailableGameSessions	Percentage of game session slots on all active server processes (healthy or unhealthy) that are not currently being used (calculated as AvailableGameSessions / [ActiveGameSessions + AvailableGameSessions + unhealthy server processes]). This metric can be used with automatic scaling. Units: Percent Relevant CloudWatch statistics: Average Dimensions: Location
GameSessionInterruptions	Number of game sessions on spot instances that have been interrupted. Units: Count Relevant CloudWatch statistics: Sum, Average, Minimum, Maximum Dimensions: Location

Player sessions

Metric	Description
CurrentPlayerSessions	Player sessions with either ACTIVE status (player is connected to an active game session) or RESERVED status (player has been given a slot in a game session but hasn't yet connected). This metric can be used with automatic scaling. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum
PlayerSessionActivations	Player sessions that transitioned from RESERVED status to ACTIVE since the last report. This occurs when a player successfully connects to an active game session. Units: Count Relevant CloudWatch statistics: Sum, Average, Minimum, Maximum

GameLift metrics for queues

The GameLift namespace includes the following metrics related to activity across a game session placement queue. Queues are used with a managed GameLift solution. The GameLift service sends metrics to CloudWatch every minute.

Metric	Description
AverageWaitTime	Average amount of time that game session placement requests in the queue with status PENDING have been waiting to be fulfilled. Units: Seconds Relevant CloudWatch statistics: Average, Minimum, Maximum, Sum Dimensions: Location
FirstChoiceNotViable	Game sessions that were successfully placed but NOT in the first-choice fleet, because that fleet was considered not viable (such as a spot fleet with a high interruption rate). This metric is based on cost, not latency. The first-choice fleet is either the first fleet listed in the queue or—when a placement request includes player latency data—it is the first fleet chosen by FleetIQ prioritization. If there are no viable spot fleets, any fleet in that region may be selected. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum, Sum
FirstChoiceOutOfCapacity	Game sessions that were successfully placed but NOT in the first-choice fleet, because that fleet had no available resources. The first-choice fleet is either the first fleet listed in the queue or—when a placement request includes player latency data —it is the first fleet chosen by FleetIQ prioritization. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum, Sum
LowestLatencyPlacement	Game sessions that were successfully placed in a region that offers the queue's lowest possible latency for the players. This metric is emitted only when player latency data is included in the placement request, which triggers FleetIQ prioritization. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum, Sum
LowestPricePlacement	Game sessions that were successfully placed in a fleet with the queue's lowest possible price for the chosen region. (FleetIQ prioritization first chooses the region with the lowest latency for the players and then finds the lowest cost fleet within that region.) This fleet can be either a spot fleet or an on-demand instance if the queue has no spot instances. This metric is emitted only when player latency data is included in the placement request. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum, Sum
Placement <region name>	Game sessions that are successfully placed in fleets located in the specified region. This metric breaks down the PlacementsSucceeded metric by region. Units: Count Relevant CloudWatch statistics: Sum
PlacementsCanceled	Game session placement requests that were canceled before timing out since the last report. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum, Sum
PlacementsFailed	Game session placement requests that failed for any reason since the last report. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum, Sum
PlacementsStarted	New game session placement requests that were added to the queue since the last report. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum, Sum
PlacementsSucceeded	Game session placement requests that resulted in a new game session since the last report. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum, Sum
PlacementsTimedOut	Game session placement requests that reached the queue's timeout limit without being fulfilled since the last report. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum, Sum
QueueDepth	Number of game session placement requests in the queue with status PENDING. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum, Sum Dimensions: Location

GameLift metrics for matchmaking

The GameLift namespace includes metrics on FlexMatch activity for matchmaking configurations and matchmaking rules. FlexMatch matchmaking is used with a managed GameLift solution. The GameLift service sends metrics to CloudWatch every minute.

For more information on the sequence of matchmaking activity, see How Amazon GameLift FlexMatch Works.

Matchmaking configurations

Metric	Description
CurrentTickets	Matchmaking requests currently being processed or waiting to be processed. Units: Count Relevant CloudWatch statistics: Average, Minimum, Maximum, Sum
MatchAcceptancesTimedOut	For matchmaking configurations that require acceptance, the potential matches that timed out during acceptance since the last report. Units: Count Relevant CloudWatch statistics: Sum
MatchesAccepted	For matchmaking configurations that require acceptance, the potential matches that were accepted since the last report. Units: Count Relevant CloudWatch statistics: Sum
MatchesCreated	Potential matches that were created since the last report. Units: Count Relevant CloudWatch statistics: Sum
MatchesPlaced	Matches that were successfully placed into a game session since the last report. Units: Count Relevant CloudWatch statistics: Sum
MatchesRejected	For matchmaking configurations that require acceptance, the potential matches that were rejected by at least one player since the last report. Units: Count Relevant CloudWatch statistics: Sum
PlayersStarted	Players in matchmaking tickets that were added since the last report. Units: Count Relevant CloudWatch statistics: Sum
TicketsFailed	Matchmaking requests that resulted in failure since the last report. Units: Count Relevant CloudWatch statistics: Sum
TicketsStarted	New matchmaking requests that were created since the last report. Units: Count Relevant CloudWatch statistics: Sum
TicketsTimedOut	Matchmaking requests that reached the timeout limit since the last report. Units: Count Relevant CloudWatch statistics: Sum
TimeToMatch	For matchmaking requests that were put into a potential match before the last report, the amount of time between ticket creation and potential match creation. Units: Seconds Relevant CloudWatch statistics: Data Samples, Average, Minimum, Maximum, p99
TimeToTicketCancel	For matchmaking requests that were canceled before the last report, the amount of time between ticket creation and cancellation. Units: Seconds Relevant CloudWatch statistics: Data Samples, Average, Minimum, Maximum, p99
TimeToTicketSuccess	For matchmaking requests that succeeded before the last report, the amount of time between ticket creation and successful match placement. Units: Seconds Relevant CloudWatch statistics: Data Samples, Average, Minimum, Maximum, p99

Matchmaking rules

Metric	Description
RuleEvaluationsPassed	Rule evaluations during the matchmaking process that passed since the last report. This metric is limited to the top 50 rules. Units: Count Relevant CloudWatch statistics: Sum
RuleEvaluationsFailed	Rule evaluations during matchmaking that failed since the last report. This metric is limited to the top 50 rules. Units: Count Relevant CloudWatch statistics: Sum

GameLift metrics for FleetIQ

The GameLift namespace includes metrics for FleetIQ game server group and game server activity as part of a FleetIQ standalone solution for game hosting. The GameLift service sends metrics to CloudWatch every minute. Also see Monitoring Your Auto Scaling Groups and Instances Using Amazon CloudWatch in the Amazon EC2 Auto Scaling User Guide.

Metric	Description
AvailableGameServers	Game servers that are available to run a game execution and are not currently occupied with gameplay. This number includes game servers that have been claimed but are still in AVAILABLE status. Units: Count Relevant CloudWatch statistics: Sum Dimensions: GameServerGroup
UtilizedGameServers	Game servers that are currently occupied with gameplay. This number includes game servers that are in UTILIZED status. Units: Count Relevant CloudWatch statistics: Sum Dimensions: GameServerGroup
DrainingAvailableGameServers	Game servers on instances scheduled for termination that are currently not supporting gameplay. These game servers are the lowest priority to be claimed in response to a new claim request. Units: Count Relevant CloudWatch statistics: Sum Dimensions: GameServerGroup
DrainingUtilizedGameServers	Game servers on instances scheduled for termination that are currently supporting gameplay. Units: Count Relevant CloudWatch statistics: Sum Dimensions: GameServerGroup
PercentUtilizedGameServers	Portion of game servers that are currently supporting game executions. This metric indicates the amount of game server capacity that is currently in use. It is useful for driving an Auto Scaling policy that can dynamically add and remove instances to match with player demand. Units: Percent Relevant CloudWatch statistics: Average, Minimum, Maximum Dimensions: GameServerGroup
GameServerInterruptions	Game servers on Spot Instances that were interrupted due to limited Spot availability. Units: Count Relevant CloudWatch statistics: Sum Dimensions: GameServerGroup, InstanceType
InstanceInterruptions	Spot Instances that were interrupted due to limited availability. Units: Count Relevant CloudWatch statistics: Sum Dimensions: GameServerGroup, InstanceType