SDK for Rust を使用した Auto Scaling の例

async fn list_groups(client: &Client) -> Result<(), Error> {
    let resp = client.describe_auto_scaling_groups().send().await?;

    println!("Groups:");

    let groups = resp.auto_scaling_groups();

    for group in groups {
        println!(
            "Name:  {}",
            group.auto_scaling_group_name().unwrap_or("Unknown")
        );
        println!(
            "Arn:   {}",
            group.auto_scaling_group_arn().unwrap_or("unknown"),
        );
        println!("Zones: {:?}", group.availability_zones(),);
        println!();
    }

    println!("Found {} group(s)", groups.len());

    Ok(())
}

[package]
name = "autoscaling-code-examples"
version = "0.1.0"
authors = ["Doug Schwartz <dougsch@amazon.com>", "David Souther <dpsouth@amazon.com>"]
edition = "2021"

# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html

[dependencies]
aws-config = { version = "1.0.1", features = ["behavior-version-latest"] }
aws-sdk-autoscaling = { version = "1.3.0" }
aws-sdk-ec2 = { version = "1.3.0" }
aws-types = { version = "1.0.1" }
tokio = { version = "1.20.1", features = ["full"] }
clap = { version = "4.4", features = ["derive"] }
tracing-subscriber = { version = "0.3.15", features = ["env-filter"] }
anyhow = "1.0.75"
tracing = "0.1.37"
tokio-stream = "0.1.14"


use std::{collections::BTreeSet, fmt::Display};

use anyhow::anyhow;
use autoscaling_code_examples::scenario::{AutoScalingScenario, ScenarioError};
use tracing::{info, warn};

async fn show_scenario_description(scenario: &AutoScalingScenario, event: &str) {
    let description = scenario.describe_scenario().await;
    info!("DescribeAutoScalingInstances: {event}\n{description}");
}

#[derive(Default, Debug)]
struct Warnings(Vec<String>);

impl Warnings {
    pub fn push(&mut self, warning: &str, error: ScenarioError) {
        let formatted = format!("{warning}: {error}");
        warn!("{formatted}");
        self.0.push(formatted);
    }

    pub fn is_empty(&self) -> bool {
        self.0.is_empty()
    }
}

impl Display for Warnings {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        writeln!(f, "Warnings:")?;
        for warning in &self.0 {
            writeln!(f, "{: >4}- {warning}", "")?;
        }
        Ok(())
    }
}

#[tokio::main]
async fn main() -> Result<(), anyhow::Error> {
    tracing_subscriber::fmt::init();

    let shared_config = aws_config::from_env().load().await;

    let mut warnings = Warnings::default();

    // 1. Create an EC2 launch template that you'll use to create an auto scaling group. Bonus: use SDK with EC2.CreateLaunchTemplate to create the launch template.
    // 2. CreateAutoScalingGroup: pass it the launch template you created in step 0. Give it min/max of 1 instance.
    // 4. EnableMetricsCollection: enable all metrics or a subset.
    let scenario = match AutoScalingScenario::prepare_scenario(&shared_config).await {
        Ok(scenario) => scenario,
        Err(errs) => {
            let err_str = errs
                .into_iter()
                .map(|e| e.to_string())
                .collect::<Vec<String>>()
                .join(", ");
            return Err(anyhow!("Failed to initialize scenario: {err_str}"));
        }
    };

    info!("Prepared autoscaling scenario:\n{scenario}");

    let stable = scenario.wait_for_stable(1).await;
    if let Err(err) = stable {
        warnings.push(
            "There was a problem while waiting for group to be stable",
            err,
        );
    }

    // 3. DescribeAutoScalingInstances: show that one instance has launched.
    show_scenario_description(
        &scenario,
        "show that the group was created and one instance has launched",
    )
    .await;

    // 5. UpdateAutoScalingGroup: update max size to 3.
    let scale_max_size = scenario.scale_max_size(3).await;
    if let Err(err) = scale_max_size {
        warnings.push("There was a problem scaling max size", err);
    }

    // 6. DescribeAutoScalingGroups: the current state of the group
    show_scenario_description(
        &scenario,
        "show the current state of the group after setting max size",
    )
    .await;

    // 7. SetDesiredCapacity: set desired capacity to 2.
    let scale_desired_capacity = scenario.scale_desired_capacity(2).await;
    if let Err(err) = scale_desired_capacity {
        warnings.push("There was a problem setting desired capacity", err);
    }

    //   Wait for a second instance to launch.
    let stable = scenario.wait_for_stable(2).await;
    if let Err(err) = stable {
        warnings.push(
            "There was a problem while waiting for group to be stable",
            err,
        );
    }

    // 8. DescribeAutoScalingInstances: show that two instances are launched.
    show_scenario_description(
        &scenario,
        "show that two instances are launched after setting desired capacity",
    )
    .await;

    let ids_before = scenario
        .list_instances()
        .await
        .map(|v| v.into_iter().collect::<BTreeSet<_>>())
        .unwrap_or_default();

    // 9. TerminateInstanceInAutoScalingGroup: terminate one of the instances in the group.
    let terminate_some_instance = scenario.terminate_some_instance().await;
    if let Err(err) = terminate_some_instance {
        warnings.push("There was a problem replacing an instance", err);
    }

    let wait_after_terminate = scenario.wait_for_stable(1).await;
    if let Err(err) = wait_after_terminate {
        warnings.push(
            "There was a problem waiting after terminating an instance",
            err,
        );
    }

    let wait_scale_up_after_terminate = scenario.wait_for_stable(2).await;
    if let Err(err) = wait_scale_up_after_terminate {
        warnings.push(
            "There was a problem waiting for scale up after terminating an instance",
            err,
        );
    }

    let ids_after = scenario
        .list_instances()
        .await
        .map(|v| v.into_iter().collect::<BTreeSet<_>>())
        .unwrap_or_default();

    let difference = ids_after.intersection(&ids_before).count();
    if !(difference == 1 && ids_before.len() == 2 && ids_after.len() == 2) {
        warnings.push(
            "Before and after set not different",
            ScenarioError::with(format!("{difference}")),
        );
    }

    // 10. DescribeScalingActivities: list the scaling activities that have occurred for the group so far.
    show_scenario_description(
        &scenario,
        "list the scaling activities that have occurred for the group so far",
    )
    .await;

    // 11. DisableMetricsCollection
    let scale_group = scenario.scale_group_to_zero().await;
    if let Err(err) = scale_group {
        warnings.push("There was a problem scaling the group to 0", err);
    }
    show_scenario_description(&scenario, "Scenario scaled to 0").await;

    // 12. DeleteAutoScalingGroup (to delete the group you must stop all instances):
    // 13. Delete LaunchTemplate.
    let clean_scenario = scenario.clean_scenario().await;
    if let Err(errs) = clean_scenario {
        for err in errs {
            warnings.push("There was a problem cleaning the scenario", err);
        }
    } else {
        info!("The scenario has been cleaned up!");
    }

    if warnings.is_empty() {
        Ok(())
    } else {
        Err(anyhow!(
            "There were warnings during scenario execution:\n{warnings}"
        ))
    }
}

pub mod scenario;


use std::{
    error::Error,
    fmt::{Debug, Display},
    time::{Duration, SystemTime},
};

use anyhow::anyhow;
use aws_config::SdkConfig;
use aws_sdk_autoscaling::{
    error::{DisplayErrorContext, ProvideErrorMetadata},
    types::{Activity, AutoScalingGroup, LaunchTemplateSpecification},
};
use aws_sdk_ec2::types::RequestLaunchTemplateData;
use tracing::trace;

const LAUNCH_TEMPLATE_NAME: &str = "SDK_Code_Examples_EC2_Autoscaling_template_from_Rust_SDK";
const AUTOSCALING_GROUP_NAME: &str = "SDK_Code_Examples_EC2_Autoscaling_Group_from_Rust_SDK";
const MAX_WAIT: Duration = Duration::from_secs(5 * 60); // Wait at most 25 seconds.
const WAIT_TIME: Duration = Duration::from_millis(500); // Wait half a second at a time.

struct Waiter {
    start: SystemTime,
    max: Duration,
}

impl Waiter {
    fn new() -> Self {
        Waiter {
            start: SystemTime::now(),
            max: MAX_WAIT,
        }
    }

    async fn sleep(&self) -> Result<(), ScenarioError> {
        if SystemTime::now()
            .duration_since(self.start)
            .unwrap_or(Duration::MAX)
            > self.max
        {
            Err(ScenarioError::with(
                "Exceeded maximum wait duration for stable group",
            ))
        } else {
            tokio::time::sleep(WAIT_TIME).await;
            Ok(())
        }
    }
}

pub struct AutoScalingScenario {
    ec2: aws_sdk_ec2::Client,
    autoscaling: aws_sdk_autoscaling::Client,
    launch_template_arn: String,
    auto_scaling_group_name: String,
}

impl Display for AutoScalingScenario {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        f.write_fmt(format_args!(
            "\tLaunch Template ID: {}\n",
            self.launch_template_arn
        ))?;
        f.write_fmt(format_args!(
            "\tScaling Group Name: {}\n",
            self.auto_scaling_group_name
        ))?;

        Ok(())
    }
}

pub struct AutoScalingScenarioDescription {
    group: Result<Vec<String>, ScenarioError>,
    instances: Result<Vec<String>, anyhow::Error>,
    activities: Result<Vec<Activity>, anyhow::Error>,
}

impl Display for AutoScalingScenarioDescription {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        writeln!(f, "\t      Group status:")?;
        match &self.group {
            Ok(groups) => {
                for status in groups {
                    writeln!(f, "\t\t- {status}")?;
                }
            }
            Err(e) => writeln!(f, "\t\t! - {e}")?,
        }
        writeln!(f, "\t         Instances:")?;
        match &self.instances {
            Ok(instances) => {
                for instance in instances {
                    writeln!(f, "\t\t- {instance}")?;
                }
            }
            Err(e) => writeln!(f, "\t\t! {e}")?,
        }

        writeln!(f, "\t        Activities:")?;
        match &self.activities {
            Ok(activities) => {
                for activity in activities {
                    writeln!(
                        f,
                        "\t\t- {} Progress: {}% Status: {:?} End: {:?}",
                        activity.cause().unwrap_or("Unknown"),
                        activity.progress.unwrap_or(-1),
                        activity.status_code(),
                        // activity.status_message().unwrap_or_default()
                        activity.end_time(),
                    )?;
                }
            }
            Err(e) => writeln!(f, "\t\t! {e}")?,
        }

        Ok(())
    }
}

#[derive(Debug)]
struct MetadataError {
    message: Option<String>,
    code: Option<String>,
}

impl MetadataError {
    fn from(err: &dyn ProvideErrorMetadata) -> Self {
        MetadataError {
            message: err.message().map(|s| s.to_string()),
            code: err.code().map(|s| s.to_string()),
        }
    }
}

impl Display for MetadataError {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        let display = match (&self.message, &self.code) {
            (None, None) => "Unknown".to_string(),
            (None, Some(code)) => format!("({code})"),
            (Some(message), None) => message.to_string(),
            (Some(message), Some(code)) => format!("{message} ({code})"),
        };
        write!(f, "{display}")
    }
}

#[derive(Debug)]
pub struct ScenarioError {
    message: String,
    context: Option<MetadataError>,
}

impl ScenarioError {
    pub fn with(message: impl Into<String>) -> Self {
        ScenarioError {
            message: message.into(),
            context: None,
        }
    }

    pub fn new(message: impl Into<String>, err: &dyn ProvideErrorMetadata) -> Self {
        ScenarioError {
            message: message.into(),
            context: Some(MetadataError::from(err)),
        }
    }
}

impl Error for ScenarioError {
    // While `Error` can capture `source` information about the underlying error, for this example
    // the ScenarioError captures the underlying information in MetadataError and treats it as a
    // single Error from this Crate. In other contexts, it may be appropriate to model the error
    // as including the SdkError as its source.
}
impl Display for ScenarioError {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        match &self.context {
            Some(c) => write!(f, "{}: {}", self.message, c),
            None => write!(f, "{}", self.message),
        }
    }
}

impl AutoScalingScenario {
    pub async fn prepare_scenario(sdk_config: &SdkConfig) -> Result<Self, Vec<ScenarioError>> {
        let ec2 = aws_sdk_ec2::Client::new(sdk_config);
        let autoscaling = aws_sdk_autoscaling::Client::new(sdk_config);

        let auto_scaling_group_name = String::from(AUTOSCALING_GROUP_NAME);

        // Before creating any resources, prepare the list of AZs
        let availablity_zones = ec2.describe_availability_zones().send().await;
        if let Err(err) = availablity_zones {
            return Err(vec![ScenarioError::new("Failed to find AZs", &err)]);
        }

        let availability_zones: Vec<String> = availablity_zones
            .unwrap()
            .availability_zones
            .unwrap_or_default()
            .iter()
            .take(3)
            .map(|z| z.zone_name.clone().unwrap())
            .collect();

        // 1. Create an EC2 launch template that you'll use to create an auto scaling group. Bonus: use SDK with EC2.CreateLaunchTemplate to create the launch template.
        //   * Recommended: InstanceType='t1.micro', ImageId='ami-0ca285d4c2cda3300'
        let create_launch_template = ec2
            .create_launch_template()
            .launch_template_name(LAUNCH_TEMPLATE_NAME)
            .launch_template_data(
                RequestLaunchTemplateData::builder()
                    .instance_type(aws_sdk_ec2::types::InstanceType::T1Micro)
                    .image_id("ami-0ca285d4c2cda3300")
                    .build(),
            )
            .send()
            .await
            .map_err(|err| vec![ScenarioError::new("Failed to create launch template", &err)])?;

        let launch_template_arn = match create_launch_template.launch_template {
            Some(launch_template) => launch_template.launch_template_id.unwrap_or_default(),
            None => {
                // Try to delete the launch template
                let _ = ec2
                    .delete_launch_template()
                    .launch_template_name(LAUNCH_TEMPLATE_NAME)
                    .send()
                    .await;
                return Err(vec![ScenarioError::with("Failed to load launch template")]);
            }
        };

        // 2. CreateAutoScalingGroup: pass it the launch template you created in step 0. Give it min/max of 1 instance.
        //   You can use EC2.describe_availability_zones() to get a list of AZs (you have to specify an AZ when you create the group).
        //   Wait for instance to launch. Use a waiter if you have one, otherwise DescribeAutoScalingInstances until LifecycleState='InService'
        if let Err(err) = autoscaling
            .create_auto_scaling_group()
            .auto_scaling_group_name(auto_scaling_group_name.as_str())
            .launch_template(
                LaunchTemplateSpecification::builder()
                    .launch_template_id(launch_template_arn.clone())
                    .version("$Latest")
                    .build(),
            )
            .max_size(1)
            .min_size(1)
            .set_availability_zones(Some(availability_zones))
            .send()
            .await
        {
            let mut errs = vec![ScenarioError::new(
                "Failed to create autoscaling group",
                &err,
            )];

            if let Err(err) = autoscaling
                .delete_auto_scaling_group()
                .auto_scaling_group_name(auto_scaling_group_name.as_str())
                .send()
                .await
            {
                errs.push(ScenarioError::new(
                    "Failed to clean up autoscaling group",
                    &err,
                ));
            }

            if let Err(err) = ec2
                .delete_launch_template()
                .launch_template_id(launch_template_arn.clone())
                .send()
                .await
            {
                errs.push(ScenarioError::new(
                    "Failed to clean up launch template",
                    &err,
                ));
            }
            return Err(errs);
        }

        let scenario = AutoScalingScenario {
            ec2,
            autoscaling: autoscaling.clone(), // Clients are cheap so cloning here to prevent a move is ok.
            auto_scaling_group_name: auto_scaling_group_name.clone(),
            launch_template_arn,
        };

        let enable_metrics_collection = autoscaling
            .enable_metrics_collection()
            .auto_scaling_group_name(auto_scaling_group_name.as_str())
            .granularity("1Minute")
            .set_metrics(Some(vec![
                String::from("GroupMinSize"),
                String::from("GroupMaxSize"),
                String::from("GroupDesiredCapacity"),
                String::from("GroupInServiceInstances"),
                String::from("GroupTotalInstances"),
            ]))
            .send()
            .await;

        match enable_metrics_collection {
            Ok(_) => Ok(scenario),
            Err(err) => {
                scenario.clean_scenario().await?;
                Err(vec![ScenarioError::new(
                    "Failed to enable metrics collections for group",
                    &err,
                )])
            }
        }
    }

    pub async fn clean_scenario(self) -> Result<(), Vec<ScenarioError>> {
        let _ = self.wait_for_no_scaling().await;
        let delete_group = self
            .autoscaling
            .delete_auto_scaling_group()
            .auto_scaling_group_name(self.auto_scaling_group_name.clone())
            .send()
            .await;

        // 14. Delete LaunchTemplate.
        let delete_launch_template = self
            .ec2
            .delete_launch_template()
            .launch_template_id(self.launch_template_arn.clone())
            .send()
            .await;

        let early_exit = match (delete_group, delete_launch_template) {
            (Ok(_), Ok(_)) => Ok(()),
            (Ok(_), Err(e)) => Err(vec![ScenarioError::new(
                "There was an error cleaning the launch template",
                &e,
            )]),
            (Err(e), Ok(_)) => Err(vec![ScenarioError::new(
                "There was an error cleaning the scale group",
                &e,
            )]),
            (Err(e1), Err(e2)) => Err(vec![
                ScenarioError::new("Multiple error cleaning the scenario Scale Group", &e1),
                ScenarioError::new("Multiple error cleaning the scenario Launch Template", &e2),
            ]),
        };

        if early_exit.is_err() {
            early_exit
        } else {
            // Wait for delete_group to finish
            let waiter = Waiter::new();
            let mut errors = Vec::<ScenarioError>::new();
            while errors.len() < 3 {
                if let Err(e) = waiter.sleep().await {
                    errors.push(e);
                    continue;
                }
                let describe_group = self
                    .autoscaling
                    .describe_auto_scaling_groups()
                    .auto_scaling_group_names(self.auto_scaling_group_name.clone())
                    .send()
                    .await;
                match describe_group {
                    Ok(group) => match group.auto_scaling_groups().first() {
                        Some(group) => {
                            if group.status() != Some("Delete in progress") {
                                errors.push(ScenarioError::with(format!(
                                    "Group in an unknown state while deleting: {}",
                                    group.status().unwrap_or("unknown error")
                                )));
                                return Err(errors);
                            }
                        }
                        None => return Ok(()),
                    },
                    Err(err) => {
                        errors.push(ScenarioError::new("Failed to describe autoscaling group during cleanup 3 times, last error", &err));
                    }
                }
                if errors.len() > 3 {
                    return Err(errors);
                }
            }
            Err(vec![ScenarioError::with(
                "Exited cleanup wait loop without retuning success or failing after three rounds",
            )])
        }
    }

    pub async fn describe_scenario(&self) -> AutoScalingScenarioDescription {
        let group = self
            .autoscaling
            .describe_auto_scaling_groups()
            .auto_scaling_group_names(self.auto_scaling_group_name.clone())
            .send()
            .await
            .map(|s| {
                s.auto_scaling_groups()
                    .iter()
                    .map(|s| {
                        format!(
                            "{}: {}",
                            s.auto_scaling_group_name().unwrap_or("Unknown"),
                            s.status().unwrap_or("Unknown")
                        )
                    })
                    .collect::<Vec<String>>()
            })
            .map_err(|e| {
                ScenarioError::new("Failed to describe auto scaling groups for scenario", &e)
            });

        let instances = self
            .list_instances()
            .await
            .map_err(|e| anyhow!("There was an error listing instances: {e}",));

        // 10. DescribeScalingActivities: list the scaling activities that have occurred for the group so far.
        //   Bonus: use CloudWatch API to get and show some metrics collected for the group.
        //   CW.ListMetrics with Namespace='AWS/AutoScaling' and Dimensions=[{'Name': 'AutoScalingGroupName', 'Value': }]
        //   CW.GetMetricStatistics with Statistics='Sum'. Start and End times must be in UTC!
        let activities = self
            .autoscaling
            .describe_scaling_activities()
            .auto_scaling_group_name(self.auto_scaling_group_name.clone())
            .into_paginator()
            .items()
            .send()
            .collect::<Result<Vec<_>, _>>()
            .await
            .map_err(|e| {
                anyhow!(
                    "There was an error retrieving scaling activities: {}",
                    DisplayErrorContext(&e)
                )
            });

        AutoScalingScenarioDescription {
            group,
            instances,
            activities,
        }
    }

    async fn get_group(&self) -> Result<AutoScalingGroup, ScenarioError> {
        let describe_auto_scaling_groups = self
            .autoscaling
            .describe_auto_scaling_groups()
            .auto_scaling_group_names(self.auto_scaling_group_name.clone())
            .send()
            .await;

        if let Err(err) = describe_auto_scaling_groups {
            return Err(ScenarioError::new(
                format!(
                    "Failed to get status of autoscaling group {}",
                    self.auto_scaling_group_name.clone()
                )
                .as_str(),
                &err,
            ));
        }

        let describe_auto_scaling_groups_output = describe_auto_scaling_groups.unwrap();
        let auto_scaling_groups = describe_auto_scaling_groups_output.auto_scaling_groups();
        let auto_scaling_group = auto_scaling_groups.first();

        if auto_scaling_group.is_none() {
            return Err(ScenarioError::with(format!(
                "Could not find autoscaling group {}",
                self.auto_scaling_group_name.clone()
            )));
        }

        Ok(auto_scaling_group.unwrap().clone())
    }

    pub async fn wait_for_no_scaling(&self) -> Result<(), ScenarioError> {
        let waiter = Waiter::new();
        let mut scaling = true;
        while scaling {
            waiter.sleep().await?;
            let describe_activities = self
                .autoscaling
                .describe_scaling_activities()
                .auto_scaling_group_name(self.auto_scaling_group_name.clone())
                .send()
                .await
                .map_err(|e| {
                    ScenarioError::new("Failed to get autoscaling activities for group", &e)
                })?;
            let activities = describe_activities.activities();
            trace!(
                "Waiting for no scaling found {} activities",
                activities.len()
            );
            scaling = activities.iter().any(|a| a.progress() < Some(100));
        }
        Ok(())
    }

    pub async fn wait_for_stable(&self, size: usize) -> Result<(), ScenarioError> {
        self.wait_for_no_scaling().await?;

        let mut group = self.get_group().await?;
        let mut count = count_group_instances(&group);

        let waiter = Waiter::new();
        while count != size {
            trace!("Waiting for stable {size} (current: {count})");
            waiter.sleep().await?;
            group = self.get_group().await?;
            count = count_group_instances(&group);
        }

        Ok(())
    }

    pub async fn list_instances(&self) -> Result<Vec<String>, ScenarioError> {
        // The direct way to list instances is by using DescribeAutoScalingGroup's instances property. However, this returns a Vec<Instance>, as opposed to a Vec<AutoScalingInstanceDetails>.
        // Ok(self.get_group().await?.instances.unwrap_or_default().map(|i| i.instance_id.clone().unwrap_or_default()).filter(|id| !id.is_empty()).collect())

        // Alternatively, and for the sake of example, DescribeAutoScalingInstances returns a list that can be filtered by the client.
        self.autoscaling
            .describe_auto_scaling_instances()
            .into_paginator()
            .items()
            .send()
            .try_collect()
            .await
            .map(|items| {
                items
                    .into_iter()
                    .filter(|i| {
                        i.auto_scaling_group_name.as_deref()
                            == Some(self.auto_scaling_group_name.as_str())
                    })
                    .map(|i| i.instance_id.unwrap_or_default())
                    .filter(|id| !id.is_empty())
                    .collect::<Vec<String>>()
            })
            .map_err(|err| ScenarioError::new("Failed to get list of auto scaling instances", &err))
    }

    pub async fn scale_min_size(&self, size: i32) -> Result<(), ScenarioError> {
        let update_group = self
            .autoscaling
            .update_auto_scaling_group()
            .auto_scaling_group_name(self.auto_scaling_group_name.clone())
            .min_size(size)
            .send()
            .await;
        if let Err(err) = update_group {
            return Err(ScenarioError::new(
                format!("Failer to update group to min size ({size}))").as_str(),
                &err,
            ));
        }
        Ok(())
    }

    pub async fn scale_max_size(&self, size: i32) -> Result<(), ScenarioError> {
        // 5. UpdateAutoScalingGroup: update max size to 3.
        let update_group = self
            .autoscaling
            .update_auto_scaling_group()
            .auto_scaling_group_name(self.auto_scaling_group_name.clone())
            .max_size(size)
            .send()
            .await;
        if let Err(err) = update_group {
            return Err(ScenarioError::new(
                format!("Failed to update group to max size ({size})").as_str(),
                &err,
            ));
        }
        Ok(())
    }

    pub async fn scale_desired_capacity(&self, capacity: i32) -> Result<(), ScenarioError> {
        // 7. SetDesiredCapacity: set desired capacity to 2.
        //   Wait for a second instance to launch.
        let update_group = self
            .autoscaling
            .set_desired_capacity()
            .auto_scaling_group_name(self.auto_scaling_group_name.clone())
            .desired_capacity(capacity)
            .send()
            .await;
        if let Err(err) = update_group {
            return Err(ScenarioError::new(
                format!("Failed to update group to desired capacity ({capacity}))").as_str(),
                &err,
            ));
        }
        Ok(())
    }

    pub async fn scale_group_to_zero(&self) -> Result<(), ScenarioError> {
        // If this fails it's fine, just means there are extra cloudwatch metrics events for the scale-down.
        let _ = self
            .autoscaling
            .disable_metrics_collection()
            .auto_scaling_group_name(self.auto_scaling_group_name.clone())
            .send()
            .await;

        // 12. DeleteAutoScalingGroup (to delete the group you must stop all instances):
        //   UpdateAutoScalingGroup with MinSize=0
        let update_group = self
            .autoscaling
            .update_auto_scaling_group()
            .auto_scaling_group_name(self.auto_scaling_group_name.clone())
            .min_size(0)
            .desired_capacity(0)
            .send()
            .await;
        if let Err(err) = update_group {
            return Err(ScenarioError::new(
                "Failed to update group for scaling down&",
                &err,
            ));
        }

        let stable = self.wait_for_stable(0).await;
        if let Err(err) = stable {
            return Err(ScenarioError::with(format!(
                "Error while waiting for group to be stable on scale down: {err}"
            )));
        }

        Ok(())
    }

    pub async fn terminate_some_instance(&self) -> Result<(), ScenarioError> {
        // Retrieve a list of instances in the auto scaling group.
        let auto_scaling_group = self.get_group().await?;
        let instances = auto_scaling_group.instances();
        // Or use other logic to find an instance to terminate.
        let instance = instances.first();
        if let Some(instance) = instance {
            let instance_id = if let Some(instance_id) = instance.instance_id() {
                instance_id
            } else {
                return Err(ScenarioError::with("Missing instance id"));
            };
            let termination = self
                .ec2
                .terminate_instances()
                .instance_ids(instance_id)
                .send()
                .await;
            if let Err(err) = termination {
                Err(ScenarioError::new(
                    "There was a problem terminating an instance",
                    &err,
                ))
            } else {
                Ok(())
            }
        } else {
            Err(ScenarioError::with("There was no instance to terminate"))
        }
    }
}

fn count_group_instances(group: &AutoScalingGroup) -> usize {
    group.instances.as_ref().map(|i| i.len()).unwrap_or(0)
}

async fn create_group(client: &Client, name: &str, id: &str) -> Result<(), Error> {
    client
        .create_auto_scaling_group()
        .auto_scaling_group_name(name)
        .instance_id(id)
        .min_size(1)
        .max_size(5)
        .send()
        .await?;

    println!("Created AutoScaling group");

    Ok(())
}

async fn delete_group(client: &Client, name: &str, force: bool) -> Result<(), Error> {
    client
        .delete_auto_scaling_group()
        .auto_scaling_group_name(name)
        .set_force_delete(if force { Some(true) } else { None })
        .send()
        .await?;

    println!("Deleted Auto Scaling group");

    Ok(())
}

async fn list_groups(client: &Client) -> Result<(), Error> {
    let resp = client.describe_auto_scaling_groups().send().await?;

    println!("Groups:");

    let groups = resp.auto_scaling_groups();

    for group in groups {
        println!(
            "Name:  {}",
            group.auto_scaling_group_name().unwrap_or("Unknown")
        );
        println!(
            "Arn:   {}",
            group.auto_scaling_group_arn().unwrap_or("unknown"),
        );
        println!("Zones: {:?}", group.availability_zones(),);
        println!();
    }

    println!("Found {} group(s)", groups.len());

    Ok(())
}

    pub async fn list_instances(&self) -> Result<Vec<String>, ScenarioError> {
        // The direct way to list instances is by using DescribeAutoScalingGroup's instances property. However, this returns a Vec<Instance>, as opposed to a Vec<AutoScalingInstanceDetails>.
        // Ok(self.get_group().await?.instances.unwrap_or_default().map(|i| i.instance_id.clone().unwrap_or_default()).filter(|id| !id.is_empty()).collect())

        // Alternatively, and for the sake of example, DescribeAutoScalingInstances returns a list that can be filtered by the client.
        self.autoscaling
            .describe_auto_scaling_instances()
            .into_paginator()
            .items()
            .send()
            .try_collect()
            .await
            .map(|items| {
                items
                    .into_iter()
                    .filter(|i| {
                        i.auto_scaling_group_name.as_deref()
                            == Some(self.auto_scaling_group_name.as_str())
                    })
                    .map(|i| i.instance_id.unwrap_or_default())
                    .filter(|id| !id.is_empty())
                    .collect::<Vec<String>>()
            })
            .map_err(|err| ScenarioError::new("Failed to get list of auto scaling instances", &err))
    }

    pub async fn describe_scenario(&self) -> AutoScalingScenarioDescription {
        let group = self
            .autoscaling
            .describe_auto_scaling_groups()
            .auto_scaling_group_names(self.auto_scaling_group_name.clone())
            .send()
            .await
            .map(|s| {
                s.auto_scaling_groups()
                    .iter()
                    .map(|s| {
                        format!(
                            "{}: {}",
                            s.auto_scaling_group_name().unwrap_or("Unknown"),
                            s.status().unwrap_or("Unknown")
                        )
                    })
                    .collect::<Vec<String>>()
            })
            .map_err(|e| {
                ScenarioError::new("Failed to describe auto scaling groups for scenario", &e)
            });

        let instances = self
            .list_instances()
            .await
            .map_err(|e| anyhow!("There was an error listing instances: {e}",));

        // 10. DescribeScalingActivities: list the scaling activities that have occurred for the group so far.
        //   Bonus: use CloudWatch API to get and show some metrics collected for the group.
        //   CW.ListMetrics with Namespace='AWS/AutoScaling' and Dimensions=[{'Name': 'AutoScalingGroupName', 'Value': }]
        //   CW.GetMetricStatistics with Statistics='Sum'. Start and End times must be in UTC!
        let activities = self
            .autoscaling
            .describe_scaling_activities()
            .auto_scaling_group_name(self.auto_scaling_group_name.clone())
            .into_paginator()
            .items()
            .send()
            .collect::<Result<Vec<_>, _>>()
            .await
            .map_err(|e| {
                anyhow!(
                    "There was an error retrieving scaling activities: {}",
                    DisplayErrorContext(&e)
                )
            });

        AutoScalingScenarioDescription {
            group,
            instances,
            activities,
        }
    }

        // If this fails it's fine, just means there are extra cloudwatch metrics events for the scale-down.
        let _ = self
            .autoscaling
            .disable_metrics_collection()
            .auto_scaling_group_name(self.auto_scaling_group_name.clone())
            .send()
            .await;

        let enable_metrics_collection = autoscaling
            .enable_metrics_collection()
            .auto_scaling_group_name(auto_scaling_group_name.as_str())
            .granularity("1Minute")
            .set_metrics(Some(vec![
                String::from("GroupMinSize"),
                String::from("GroupMaxSize"),
                String::from("GroupDesiredCapacity"),
                String::from("GroupInServiceInstances"),
                String::from("GroupTotalInstances"),
            ]))
            .send()
            .await;

    pub async fn scale_desired_capacity(&self, capacity: i32) -> Result<(), ScenarioError> {
        // 7. SetDesiredCapacity: set desired capacity to 2.
        //   Wait for a second instance to launch.
        let update_group = self
            .autoscaling
            .set_desired_capacity()
            .auto_scaling_group_name(self.auto_scaling_group_name.clone())
            .desired_capacity(capacity)
            .send()
            .await;
        if let Err(err) = update_group {
            return Err(ScenarioError::new(
                format!("Failed to update group to desired capacity ({capacity}))").as_str(),
                &err,
            ));
        }
        Ok(())
    }

    pub async fn terminate_some_instance(&self) -> Result<(), ScenarioError> {
        // Retrieve a list of instances in the auto scaling group.
        let auto_scaling_group = self.get_group().await?;
        let instances = auto_scaling_group.instances();
        // Or use other logic to find an instance to terminate.
        let instance = instances.first();
        if let Some(instance) = instance {
            let instance_id = if let Some(instance_id) = instance.instance_id() {
                instance_id
            } else {
                return Err(ScenarioError::with("Missing instance id"));
            };
            let termination = self
                .ec2
                .terminate_instances()
                .instance_ids(instance_id)
                .send()
                .await;
            if let Err(err) = termination {
                Err(ScenarioError::new(
                    "There was a problem terminating an instance",
                    &err,
                ))
            } else {
                Ok(())
            }
        } else {
            Err(ScenarioError::with("There was no instance to terminate"))
        }
    }

    async fn get_group(&self) -> Result<AutoScalingGroup, ScenarioError> {
        let describe_auto_scaling_groups = self
            .autoscaling
            .describe_auto_scaling_groups()
            .auto_scaling_group_names(self.auto_scaling_group_name.clone())
            .send()
            .await;

        if let Err(err) = describe_auto_scaling_groups {
            return Err(ScenarioError::new(
                format!(
                    "Failed to get status of autoscaling group {}",
                    self.auto_scaling_group_name.clone()
                )
                .as_str(),
                &err,
            ));
        }

        let describe_auto_scaling_groups_output = describe_auto_scaling_groups.unwrap();
        let auto_scaling_groups = describe_auto_scaling_groups_output.auto_scaling_groups();
        let auto_scaling_group = auto_scaling_groups.first();

        if auto_scaling_group.is_none() {
            return Err(ScenarioError::with(format!(
                "Could not find autoscaling group {}",
                self.auto_scaling_group_name.clone()
            )));
        }

        Ok(auto_scaling_group.unwrap().clone())
    }

async fn update_group(client: &Client, name: &str, size: i32) -> Result<(), Error> {
    client
        .update_auto_scaling_group()
        .auto_scaling_group_name(name)
        .max_size(size)
        .send()
        .await?;

    println!("Updated AutoScaling group");

    Ok(())
}