Create event wizard, change requests, instance managers and action logs

[timalces]

Aims to resolve #4

Creation Wizard

• Adds a new page Create Events, for users to schedule turning instances on and/or off
• Uses an on page 'wizard', with multiple sections shown sequentially (without page reload)
• Users can select either minimum counts or exact counts, by group and instance type. Minimum counts will only involve switch ons, whereas exact counts may involve switch ons and/or switch offs
• Users can select either to carry out the change now or in the future. If now, it will be carried out 5 minutes (rounded up to nearest whole minute) after the request is submitted
• If in the future, the user can select the date and time. They may also optionally choose to repeat the request on the chosen days of the week, until the chosen end date
• If in the future, the user may also optionally choose a description for the request
• Once all choices have been made, the final wizard page shows a simplified cost chart showing the projected costs based on the request
• This will highlight if the request is forecast to go over budget or over balance within the current billing cycle. In either situation the user will be prevented from submitting the request

Change Requests

• On submitting a valid request, a ChangeRequest object will be created
• This request will either be a OneOffChangeRequest or a RepeatedChangeRequest
• These requests will be used when calculating costs forecasts, both in the Costs Breakdown page and the event wizard
• A slack message is also sent when a ChangeRequest is created via the wizard

• This currently always says 'someone' submitted the request, but in future will name the user (once users are implemented)

Carrying out Change Requests

• A new cron task is added (via whenever) that checks if any Change Requests are due, each minute
• If so, the request's targets are compared against the current counts. If any differences, the required actions & instances to change are determined and submitted to the Azure API/ AWS SDK, via new service classes AzureInstanceManager and AwsInstanceManager. These then handle switching instances on/off
• When this happens ActionLogs are also created, recording details of the instance status change. These action logs are also used when calculating cost forecasts, to estimate instances' up-time (on forecast days in the past, and today)
• If counts are already met, no actions will be taken
• A slack message with details is also sent when a change request is carried out:

With Actions:

No Actions:

Manual Action Logs

• If an instance is switched on or off from outside the application, an action log must be created recording the event for cost forecasts to work correctly (and prevent unexpected behaviour)
• This can be created using using the task rake action_logs:add [project_name,action,reason,instance_id,actioned_at_time].
• instance_id here refers to the id given by the instance's platform and must be in the appropriate format. For example, for aws this would be something like i-0b00efe3aab7010da, or for azure /subscriptions/#{subscription_id}/resourceGroups/#{resource_group}/providers/Microsoft.Compute/virtualMachines/#{instance_name}.
• action must be on or off and actioned_at_time in the format "yyyy-mm-dd HH:MM".

[ColonelPanicks]

I have an event that will ensure node counts are at 0 at 2am on 24/2/22

Someone requested the following scheduled exact counts for **democluster**:
**default**
Standard_B2s: 0 nodes
Standard_DS1_v2: 0 nodes
**special**
Standard_F2s_v2: 0 nodes
Standard_B2s: 0 nodes
**Scheduled time:** 2022-02-24 02:00:00 +0000

Yet the UI seems to be assuming the 1 node that is up will be up all of tomorrow. Is this expected? I'd have expected to see ~3 compute units as the estimated cost for the node tomorrow.

[timalces]

Merging this as appears to be working as intended.