Set an elastic policy

View as Markdown

An elastic policy adds one node when CPU use is above the value you set, and removes one node when it is below a second, lower value. The policy keeps the scale group between its minimum and maximum.

You set the policy when you create the scale group, with Elastic Policy or Elastic and Scheduled Policy under Select Policy, and you can change it later on the scale group’s Scale Group Details tab. To scale at set times instead, see Set a scheduled policy.

Before you begin

For a new scale group, you have chosen Elastic Policy or Elastic and Scheduled Policy under Select Policy while creating it, and your role in the current project has the create permission for Auto Scaling. See Create a scale group.
If you are changing an existing scale group, it reads Running, and your role has the update permission for Auto Scaling.

Choose the parameter

Keep Policy Parameter Type on Default, and pick CPU under Select a policy parameter. Default offers only CPU, so the policy scales on CPU use alone. To scale on memory or another measurement, watch it with your own monitoring and set the count with Resize Service, or from a script with the as-resize operation; see Resize the scale group.

Set when a node is added and removed

The policy has two rules: Add node when, the scale-up rule, and Remove node when, the scale-down rule. Each time a rule fires, the scale group adds or removes one node. For Add node when, choose more than or more than equal to. For Remove node when, choose less than or less than equal to. To make a rule fire at the value itself too, choose more than equal to or less than equal to.

Both values are percentages:

Add node when takes a value from 30 to 99, and starts at 60 with more than.
Remove node when takes a value from 10 up to half the scale-up value, and starts at 30 with less than. When you type a scale-up value, the portal sets Remove node when to half of it, rounded down, and you can lower it from there.

The CPU percentage is the node’s CPU use as a share of its plan’s vCPUs. For example, 60% on a four-vCPU plan means 2.4 vCPUs are busy.

Set the watch period and cooldown

The scale group adds a node when CPU use stays above the scale-up value for Watch Period periods in a row, each Period durations(s) long. It removes a node the same way, when CPU use stays below Remove node when for that many periods. After each scaling action, the scale group waits for Cooldown(s) before it scales again.

Watch Period is the number of periods in a row. It starts at 3.
Period durations(s) is the length of each period, in seconds. It starts at 10.
Cooldown(s) is the wait after a scaling action, in seconds. It starts at 150.

Each takes any whole number from zero up. Under the fields, the portal sums up the policy in one sentence with your values in it. Read it before you save.

With the starting values, once a steady load goes above the scale-up value, the scale group starts adding a node after about 4.5 minutes, and the node takes about 2.5 minutes more to be added. Removing a node takes about 20 seconds.

Change the policy of a running scale group

In MyAccount, under Compute, click Auto Scaling, click your scale group, and open its Scale Group Details tab. You edit its settings there one at a time.
Click the edit button beside Scale group policy. Select Policy opens with the policy’s values ready to change.
Choose Elastic Policy, Scheduled Policy, or Elastic and Scheduled Policy, and change the values, the watch period, the period duration, and the cooldown.
Click Update to save, or Cancel to drop the changes. Saving shows this message, where <group-name> is your scale group’s name:
```
The scaling service - <group-name> - has been successfully updated.
```

The Activity Timeline records the change as UPDATE_POLICY.

Check the policy

The elastic policy’s panel on Scale Group Details shows Policy Parameter and the two rules, under Adds a node when and Removes a node when. It also shows Watch period policy, Duration (second), and Cooldown (second), as you set them.

When the policy adds or removes a node, the scale group’s Audit Logs tab records it; see Read the audit logs. To see the nodes running now, open Active Node Details.

Your scale group now scales on CPU use. To also scale at set times, see Set a scheduled policy. To remove every node for a while, see Stop the scale group. If the scale group scales differently from what you expect, see Troubleshooting Auto Scaling.