-
Notifications
You must be signed in to change notification settings - Fork 3
Cloudwatch alarms
The cloudformation defines alarms for monitoring. When fired, the alarms ping a webhook specified as the AlertWebhook parameter, allowing integration with tools like Pager Duty.
Fires if any 5XX responses have been seen within a 5 minute period.
Original Problem: Issues with auth and the internal Dynamo schemas
Rationale: It's a good idea to investigate every unexpected exception!
If such alerts become a regular occurrence and are ignored the alarm should be updated to be less trigger happy.
Fires on repeated occurrence of 4XX responses. This is defined as more than 15 in a ten minute period.
Rationale: Repeated client error responses probably indicate a bug with a client.
The alarm criteria was somewhat arbitrarily copied from The Grid and should be reviewed based on application usage over time.
## MediaAtomMakerLatency
Fires if requests start taking more than 1 second on average across a 5 minute period.
Rationale: Unexpectedly slow requests need investigation!
## MediaAtomMAkerUnhealthy
Fires if any of the instances behind the ELB are considered unhealthy.
Original problem: The application went down without firing any of the existing alarms, because they were all request-based.
Rationale: Quick raw indicator of degradation of service (not necessarily full failure)