Monitor and roll back stack operations
Rollback triggers enable you to have AWS CloudFormation monitor the state of your
application during
stack creation and updating, and to roll back that operation if the application breaches
the
threshold of any of the alarms you've specified. For each rollback trigger you create,
you
specify the CloudWatch alarm that CloudFormation should monitor. CloudFormation monitors
the specified alarms
during the stack create or update operation, and for the specified amount of time
after all
resources have been deployed. If any of the alarms goes to ALARM state during the stack
operation or the monitoring period, CloudFormation rolls back the entire stack operation.
You can set a monitoring time from the default of 0 up to 180 minutes. During this
time,
CloudFormation monitors all the rollback triggers after the stack creation or update
operation
deploys all necessary resources. If any of the alarms goes to ALARM state during the stack
operation or this monitoring period, CloudFormation rolls back the entire stack operation.
Then,
for update operations, if the monitoring period expires without any alarms going to
ALARM
state, CloudFormation proceeds to dispose of old resources as usual. If you set a
monitoring
time but don't specify any rollback triggers, CloudFormation still waits the specified
period of
time before cleaning up old resources for update operations. You can use this monitoring
period to perform any manual stack validation desired, and manually cancel the stack
creation or update as necessary. If you set a monitoring time of 0 minutes, CloudFormation
still
monitors the rollback triggers during stack creation and update operations and rolls
back
the operation if an alarm goes to ALARM state. Then, for update operations with no breaching
alarms, it begins disposing of old resources immediately once the operation
completes.
By default, CloudFormation only rolls back stack operations if an alarm goes to ALARM
state, not INSUFFICIENT_DATA state. To have CloudFormation roll back the stack operation if an alarm
goes to INSUFFICIENT_DATA state as well, edit the CloudWatch alarm to treat missing data as
breaching. For more information, see Configuring how
CloudWatch alarms treat missing data in Amazon CloudWatch User Guide.
CloudFormation doesn't monitor rollback triggers when it rolls back a stack during an update operation.
You can add a maximum of five rollback triggers. To add a rollback trigger, you specify the ARN (Amazon Resource Name) of the CloudWatch alarm. Currently, only AWS::CloudWatch::Alarm types can be used as rollback triggers.
If a given CloudWatch alarm is missing, the entire stack operation fails and is rolled back.
Be aware that access to Amazon CloudWatch requires credentials. Those credentials must have permissions to access AWS resources, such as retrieving CloudWatch metric data about your cloud resources. For more information, see Authentication and access control for Amazon CloudWatch in Amazon CloudWatch User Guide.
To add rollback triggers during stack creation or updating
-
During creating or updating a stack, on the Configure stack options page, under Advanced options, go to Rollback configuration.
-
Specify a monitoring time between 0–180 minutes. The default value is 0.
-
Enter the ARN of the Cloudwatch alarm you want to use as a rollback trigger, and click Add CloudWatch alarm ARN.
You can add a maximum of five rollback triggers.
To add rollback triggers to a change set
-
During creating or updating a change set, on the Configure stack options page, under Advanced options, go to Rollback configuration.
-
Specify a monitoring time between 0–180 minutes. The default value is 0.
-
Enter the ARN of the Cloudwatch alarm you want to use as a rollback trigger, and click Add CloudWatch alarm ARN.
You can add a maximum of five rollback triggers.
To view rollback triggers for a stack
-
On the Stacks page, select the stack you wish to view from the list on the left. On the Stack info tab, under Advanced options, expand the Rollback configuration section.
