put metric action

[kapilt]

for given expression evaluate and publish per resource metric for given attribute, or the collection's aggregate attributes (size, etc).

[calvinhp]

Are you envisioning something like this:

policies:
  - name: track-bucket-size
    resource: s3
    comment: |
      Put the sum of the size of all keys in an S3 bucket into CloudWatch Metrics
    filter:
      - Name: metric-tracked-bucket
    actions:
      - type: put-metric
        key: Contents[].Size
        metric_name: Bucket Size
        op: sum

Or something like this:

  - name: track-attached-ebs
    resource: ec2
    comment: |
      Put the count of the number of EBS attached disks to an instance
    filter:
      - Name: tracked-ec2-instance
    actions:
      - type: put-metric
        key: Reservations[].Instances[].BlockDeviceMappings[].DeviceName
        metric_name: Attached Disks
        op: count

The action can figure out based on the resource, operation and the key handed to it what to put into CloudWatch.

Operations could be:

• count
• sum
• average
• duration (is this even possible?)

[kapilt]

That looks good to me, thanks

[gfranxman]

Count, Sum and Average are definitely doable as Operations. Duration is really a question of the units for the value being recorded. I think the right thing to do here is to add one more key to the action calls 'units' and accept values of
`

durations

'Seconds','Microseconds','Milliseconds',

bytes

'Bytes','Kilobytes','Megabytes','Gigabytes','Terabytes',

bits

'Bits','Kilobits','Megabits','Gigabits','Terabits',

rates

'Bytes/Second','Kilobytes/Second','Megabytes/Second','Gigabytes/Second','Terabytes/Second'

'Bits/Second','Kilobits/Second','Megabits/Second','Gigabits/Second','Terabits/Second'

'Count/Second',

other

'Percent',
'Count',
'None'

`
This offers a way to interpret the values that graph as lines or area charts, or as values in a table.

Cloud watch can also record statistical values at a point in time. In that case, rather than a value, it accepts statistical metadata about a sample set ( min, max, sum, and sample count ). I've not used that yet, but we could support that style as an additional op value called 'stats'.