Skip to main content

Starting, stopping and rebooting

Each resource on the Resources page has a power switch in the Power column. Several resources can be started, stopped or rebooted at once from the bar at the bottom of the page.

What you may do is set by your access rules: Start, Stop and Reboot are separate rights. Delegaris checks them against the resource's tags in AWS at the moment you act, not against what your browser showed.

The power switch​

The switch is on (green) when the resource is running and off when it is stopped. Hover over it to see what it will do, for example "Stop web-1".

  • Starting: click a switch that is off. The resource starts straight away, with no question asked.
  • Stopping: click a switch that is on. A dialog asks you to confirm (see below).

The switch is greyed out when you do not have the right; hovering over it says "You do not have permission to stop web-1". It is also greyed out while the resource is changing state.

When you act, a message in the corner of the page says what happened: "Start requested for web-1", or "Could not start web-1" with the reason. AWS takes a little while to carry out a start or stop; the list refreshes every few seconds until it has.

To start or stop several resources, tick them and use Start or Stop in the bar at the bottom. Each button counts only the selected resources it applies to.

The stop dialog​

The dialog lists what will be stopped, with each resource's type, account and region, and adds a warning where one applies:

WarningWhen
"...belongs to an Auto Scaling group. Auto Scaling usually terminates and replaces a stopped instance instead of leaving it stopped."An EC2 instance that is part of an Auto Scaling group. Stop the group instead.
"...is being kept on. Stopping now is fine: the schedule tag still goes back at the time that was set."The resource is being kept on.
"Its size is set to zero. The size before is remembered, and Start puts it back."Auto Scaling groups, ECS services and EKS node groups.
"AWS starts a stopped database again by itself after seven days."RDS databases and clusters.

Press Stop name to go ahead, or Cancel.

Busy states​

While a resource is starting, stopping or otherwise changing, its dot and switch show it as changing and the switch cannot be used. Hovering over the switch says what it is doing, for example "web-1 is stopping". The Changing filter above the list shows everything in this state.

Starting something that is already running, or stopping something already stopped, does nothing and says "Already running" (or stopped). Starting or stopping something mid-change is refused with "Cannot stop while the resource is starting"; wait for it to finish.

Rebooting​

Reboot restarts a running resource in place, without stopping it: for the moment someone asks for a machine to be turned off and on again.

  • Open the resource and press Reboot in its detail panel, or tick several and press Reboot in the bar. Either way you are asked to confirm.
  • It only appears for running resources, and only if your access rule includes Reboot. It is a separate right from start and stop.

Not every kind of resource can be rebooted:

KindReboot
EC2 instances, RDS instances, Redshift clusters, Lightsail instancesRestart in place
ECS servicesEvery task is replaced (a new deployment)
Aurora, DocumentDB and Neptune clustersNot offered; reboot member instances one at a time in AWS
Auto Scaling groups, EKS node groups, SageMaker notebooksNot offered; stop and start instead
Administrators

Reboot needs ec2:RebootInstances and rds:RebootDBInstance in the member account role. If you deployed the role before reboot existed, redeploy it. See Member account role.

Scaled services​

Auto Scaling groups, ECS services and EKS node groups have no "stopped" state of their own. Delegaris stops them by setting their size to zero, and starts them by putting the size back:

ServiceStopStart
Auto Scaling groupsMinimum and desired set to 0Back to the minimum, desired and maximum it had
ECS servicesDesired count set to 0Back to the count it had
EKS managed node groupsMinimum and desired set to 0Back to the size it had

The size before the stop is kept in Delegaris's own database, not in a tag, so nobody can edit it away. While one of these is stopped, its details show Starts at with the size a start will bring back. If it was stopped some other way and Delegaris has no size remembered, a start brings it up to one.

ECS services with Application Auto Scaling

If an ECS service's task count is managed by Application Auto Scaling with a minimum above zero, Auto Scaling scales it straight back up. Set that minimum to 0, or suspend scaling, for services you stop here.

See Supported services for every kind of resource and what start and stop mean for each.

Actions that need approval​

An access rule can say that its actions need approval first. When it does:

  • Buttons for those actions show (ask), for example Keep on… (ask) or Reboot (ask). The power switch looks the same as usual.
  • Pressing it does nothing to the resource. A request goes to the approvers instead, and a message says "Sent for approval", with "Nothing happens until an approver agrees. You can follow it under Approvals."
  • The resource shows Waiting for approval in its Tags column until the request is decided. Click it to open Approvals.
  • Once an approver agrees, the action runs straight away, as you.

If another rule gives you the same action outright, that rule wins and nothing is asked. See Approvals.

Protected resources​

A protected resource can be seen by whoever a rule shows it to, but only administrators can start, stop, reboot, keep it on or change its tags. Its detail panel says so.

What is recorded​

Every start, stop and reboot, and every refusal, is recorded under Activity with who did it and from which address. In accounts reached through a role, the AWS call is made under the role session delegaris-<username>, so CloudTrail shows the person too.