Recent AWS Customer Success Stories & Videos

More AWS Customer Success Stories...

« AWS Week in Review - December 31, 2012 | Main | Annoucing the Finalists of the 2012 AWS Start-Up Challenge »

TrackBack

TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00d8341c534853ef017ee6296f34970d

Listed below are links to weblogs that reference Amazon CloudWatch - Alarm Actions:

Comments

Feed You can follow this conversation by subscribing to the comment feed for this post.

Thefalken

Awesome work, I'd love plain old 'restart' as an action too.

Steven

In recent weeks some of our EC2 instances have been hit with underlying hardware issues and they restart automatically. Is it planned this this feature could be extended to not only stop but start an instance and re-assign the static EIP if for example a system or instance check fails and is caught by a CloudWatch alarm?

Yaron

The AWS CloudWatch CLI (http://aws.amazon.com/developertools/2534) has not been updated since September. When can we expect it to support this new, and very useful, functionality?

Thanks!

JB

Cool, sounds like it's getting closer and closer to some of the things I've asked for in the AWS forums long ago.

It would be nice to also be able to use the *current time* as a threshold or criteria of an action. Such as stopping an unutilized server during the night, and starting it up again in the morning (This would also require reassigning the EIP to the instance).

Or during a certain *window of time* and low utilization, I would like to have it take a snapshot or create an AMI of the instance for me.

For now, I've had to build a service which runs on another server to coordinate all of this, but I'm sure it's not as reliable as it would be if this type of service was built into CloudWatch.

Bjorn

How can this be applied to all instances instead of only a specific one?

Derek

-- @Yaron,
The existing Amazon CloudWatch Command-Line Tools (CLI) do support these new actions. In the "--alarm-actions" parameter, specify an Amazon Resource Name that corresponds to the action you want to use (remember to include the region where your instance runs). For example, to stop an instance in us-east-1, use "arn:aws:automate:us-east-1:ec2:stop" or to terminate that instance, use "arn:aws:automate:us-east-1:ec2:terminate". You can find more information in the Amazon CloudWatch developer guide at: http://docs.amazonwebservices.com/AmazonCloudWatch/latest/DeveloperGuide/UsingAlarmActions.html

-- @Steven and @Thefalken,
Thanks for the feedback about your desire to see reboot and stop/start actions. Currently you can use this feature to stop or terminate an instance that fails a status check since Amazon CloudWatch monitors status check results. However, we hear you loud and clear that re-starting the instance (and reassigning the Elastic IP address) would be even more useful.

-- @JB,
Thanks for the suggestions for time-based stopping and actions such as snapshot or AMI creation. You may want to consider Auto Scaling, which has a scheduled scaling option (using the PutScheduledUpdateGroupAction API); using that, you can terminate an instance each night and re-launch it each morning. http://docs.aws.amazon.com/AutoScaling/latest/DeveloperGuide/Welcome.html

-- @Bjorn,
We currently do not have simple methods to apply this to multiple instances at once, but you can use tools such as the command-line interface for this. Or, you may be able to use the EC2 console; it remembers your prior alarm's settings so you can go through your list of instances and create identical alarms for multiple instances in a few seconds each.

Nick

+1 to a reboot action. I can't even think of a case where we would just shut it down...

DR Shaw

An out of band monitoring option would be nice as well. I currently use cloudwatch and external monitoring to ensure I have out of band monitoring. I need to ensure I measure user experience and not just resource monitoring. Cloudwatch is susceptible to AWS resource outages as well.

Dan Mackin

This looks awesome! From the forms I found this: "Or, if you are a corporate IT administrator, you can create a group of alarms that first sends an email notification to developers whose instances have been underutilized for a day, then stops an instance if utilization doesn't improve after three days, and terminates the instance after a week of no activity."

However, when trying to setup an alarm for stopping an instance after three days of under-utilization, I'm getting the following error: "The number of consecutive periods cannot span more than a day."

Any ideas? These seem contradictory... is monitoring for longer than a day not yet supported? Thanks!
-Dan

The comments to this entry are closed.

Featured Events

The AWS Report


Brought to You By

Jeff Barr (@jeffbarr):



Jinesh Varia (@jinman):


Email Subscription

Enter your email address:

Delivered by FeedBurner

April 2014

Sun Mon Tue Wed Thu Fri Sat
    1 2 3 4 5
6 7 8 9 10 11 12
13 14 15 16 17 18 19
20 21 22 23 24 25 26
27 28 29 30