search

caltechads / deployfish

Manage your whole application lifecycle in Amazon AWS ECS.
http://deployfish.readthedocs.io
Other
90 stars 20 forks source link

Autoscaling Not Working #38

Closed thestephenstanton closed 5 years ago

thestephenstanton commented 5 years ago

https://github.com/caltechads/deployfish/blob/57c23bc7b4a267b3bc4fcf1c6288e2d71d89afbd/deployfish/aws/test/test_ECSServiceCPUAlarm.py#L78

In the line above, it seems to be causing Cloudwatch's alarm to never work.

So what exactly happens. When creating the autoscaling policy using deployfish, it posts the json to aws just fine; but the state of the alarm never changes from INSUFFICIENT_DATA. If you go in the aws console and select that alarm and then just hit save on it, state of the alarm will go to OK.

The reason seems to be because "Unit" gets posted to aws which is no longer a valid field

cmalek commented 5 years ago

Egads, not good. Thanks for pointing this out. I'll have a fix out shortly.

Chris

cmalek commented 5 years ago

Ah, it's not that the 'Unit' field itself is invalid, it's that the value i'm passing is wrong for cpu-- it should be 'Percent'. I'll fix.

thestephenstanton commented 5 years ago

Ah ok, gotcha. Good find then.

On Wed, Apr 17, 2019 at 3:35 PM Chris Malek notifications@github.com wrote:

Ah, it's not that the 'Unit' field itself is invalid, it's that the value i'm passing is wrong for cpu-- it should be 'Percent'. I'll fix.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/caltechads/deployfish/issues/38#issuecomment-484231052, or mute the thread https://github.com/notifications/unsubscribe-auth/AHJZ3MEzLLP75eJEVZFijGbEc5NEy-3pks5vh3edgaJpZM4c1yBr .

cmalek commented 5 years ago

This is fixed. New alarms should be created correctly, and If you do deploy update <service_name>, that will fix your existing service alarms.

This is on PyPI as deployfish-0.28.1.