Allszn's picture
Add launch script
a399cf4 verified
Raw History Blame Contribute Delete
11 kB
#!/usr/bin/env bash
#
# launch.sh - deploy the Windows GPU cloud-gaming host.
#
# Validates the launch kit, creates the S3 artifact bucket, deploys the
# CloudFormation stack, waits for the instance to register with Systems Manager,
# then pushes the Windows bootstrap script over SSM. No inbound RDP, ever.
#
# Quick start:
# export AWS_PROFILE=my-profile KEY_NAME=my-keypair
# ./launch.sh --instance-type g6.xlarge --volume-gib 600
#
set -euo pipefail
# ----------------------------------------------------------------- defaults
STACK_NAME="${STACK_NAME:-steam-gpu}"
REGION="${AWS_REGION:-${AWS_DEFAULT_REGION:-us-east-1}}"
INSTANCE_TYPE="${INSTANCE_TYPE:-g6.xlarge}"
GAME_VOLUME_GIB="${GAME_VOLUME_GIB:-600}"
ADMIN_CIDR="${ADMIN_CIDR:-}"
KEY_NAME="${KEY_NAME:-}"
ARTIFACT_BUCKET="${ARTIFACT_BUCKET:-}"
BOOTSTRAP="${BOOTSTRAP:-1}"
DRY_RUN="${DRY_RUN:-0}"
TEMPLATE="cloudformation/steam-gpu-host.yaml"
BOOTSTRAP_PS1="scripts/bootstrap.ps1"
DIAGNOSE_PS1="scripts/diagnose.ps1"
SSM_DOC="ssm/SteamGpu-Diagnose.json"
# ------------------------------------------------------------------- output
if [[ -t 1 ]]; then
C_B=$'\033[1;34m'; C_Y=$'\033[1;33m'; C_R=$'\033[1;31m'; C_G=$'\033[1;32m'; C_0=$'\033[0m'
else
C_B=''; C_Y=''; C_R=''; C_G=''; C_0=''
fi
log() { printf '%s[launch]%s %s\n' "$C_B" "$C_0" "$*"; }
ok() { printf '%s[ ok ]%s %s\n' "$C_G" "$C_0" "$*"; }
warn() { printf '%s[warn]%s %s\n' "$C_Y" "$C_0" "$*" >&2; }
die() { printf '%s[fail]%s %s\n' "$C_R" "$C_0" "$*" >&2; exit 1; }
usage() {
sed -n '2,20p' "$0" | sed 's/^# \{0,1\}//'
cat <<'EOF'
Options:
--stack-name NAME CloudFormation stack name (default: steam-gpu)
--region REGION AWS region (default: us-east-1)
--instance-type TYPE g4dn.xlarge|g5.xlarge|g6.xlarge|... (default: g6.xlarge)
--key-name NAME EC2 key pair (decrypts the local Administrator password)
--admin-cidr CIDR CIDR allowed to reach DCV, e.g. 1.2.3.4/32
(default: your current public IP)
--volume-gib N Steam library EBS size in GiB (default: 600)
--artifact-bucket NAME Bucket for bootstrap scripts (default: created)
--no-bootstrap Create infrastructure only; skip the SSM bootstrap
--dry-run Validate and print the plan; change nothing
-h, --help This text
Every option is also settable as the uppercase environment variable of the same name:
STACK_NAME AWS_REGION INSTANCE_TYPE KEY_NAME ADMIN_CIDR GAME_VOLUME_GIB
ARTIFACT_BUCKET DRY_RUN
EOF
exit 0
}
# --------------------------------------------------------------------- args
while [[ $# -gt 0 ]]; do
case "$1" in
--stack-name) STACK_NAME="$2"; shift 2 ;;
--region) REGION="$2"; shift 2 ;;
--instance-type) INSTANCE_TYPE="$2"; shift 2 ;;
--key-name) KEY_NAME="$2"; shift 2 ;;
--admin-cidr) ADMIN_CIDR="$2"; shift 2 ;;
--volume-gib) GAME_VOLUME_GIB="$2"; shift 2 ;;
--artifact-bucket) ARTIFACT_BUCKET="$2"; shift 2 ;;
--no-bootstrap) BOOTSTRAP=0; shift ;;
--dry-run) DRY_RUN=1; shift ;;
-h|--help) usage ;;
*) die "unknown option: $1 (try --help)" ;;
esac
done
# ------------------------------------------------------------ preconditions
command -v aws >/dev/null 2>&1 || die "aws CLI not found"
command -v python3 >/dev/null 2>&1 || die "python3 not found (needed by the validator)"
[[ -f "$TEMPLATE" ]] || die "missing $TEMPLATE - run this from the launch kit root"
aws sts get-caller-identity >/dev/null 2>&1 \
|| die "no usable AWS credentials. Set AWS_PROFILE or AWS_ACCESS_KEY_ID/SECRET."
ACCOUNT="$(aws sts get-caller-identity --query Account --output text)"
CALLER="$(aws sts get-caller-identity --query Arn --output text)"
log "account ${ACCOUNT} region ${REGION} stack ${STACK_NAME}"
log "caller ${CALLER}"
if [[ -z "$ADMIN_CIDR" ]]; then
MYIP="$(curl -fsS https://checkip.amazonaws.com | tr -d '\n')" \
|| die "could not determine your public IP; pass --admin-cidr explicitly"
ADMIN_CIDR="${MYIP}/32"
warn "defaulted --admin-cidr to ${ADMIN_CIDR}. Your IP changes; re-run or update the rule later."
fi
[[ "$ADMIN_CIDR" =~ /32$ ]] || die "--admin-cidr must be a single host, e.g. 203.0.113.7/32"
# ---------------------------------------------------------------- step 1: validate
log "step 1/5 - validating the launch kit (no AWS calls)"
PY=python3
$PY -c 'import yaml' 2>/dev/null || {
warn "pyyaml missing; attempting a local install"
$PY -m pip install --quiet --user pyyaml || die "install pyyaml to validate: python3 -m pip install pyyaml"
}
if $PY scripts/validate.py --root . --pretty; then
ok "validation passed"
else
die "validation failed - fix the checks above before deploying"
fi
# ---------------------------------------------------------------- step 2: bucket
if [[ -z "$ARTIFACT_BUCKET" ]]; then
ARTIFACT_BUCKET="steam-gpu-launchkit-${ACCOUNT}-${REGION}"
log "step 2/5 - no --artifact-bucket given; will use ${ARTIFACT_BUCKET}"
else
log "step 2/5 - using artifact bucket ${ARTIFACT_BUCKET}"
fi
# ---------------------------------------------------------------- step 3: deploy
DEPLOY_CMD=(
aws cloudformation deploy
--stack-name "$STACK_NAME"
--template-file "$TEMPLATE"
--capabilities CAPABILITY_NAMED_IAM
--no-fail-on-empty-changeset
--region "$REGION"
--parameter-overrides
"InstanceType=${INSTANCE_TYPE}"
"KeyName=${KEY_NAME}"
"AdminCidr=${ADMIN_CIDR}"
"ArtifactBucket=${ARTIFACT_BUCKET}"
"GameVolumeSizeGiB=${GAME_VOLUME_GIB}"
)
log "step 3/5 - deploying the stack"
printf ' %s\n' "${DEPLOY_CMD[*]}"
if [[ "$DRY_RUN" == "1" ]]; then
warn "dry run: stopping before any AWS change."
ok "plan printed. Re-run without --dry-run to apply."
exit 0
fi
[[ -n "$KEY_NAME" ]] || die "KEY_NAME/--key-name is required for a real deploy"
# Create the artifact bucket first: the instance role's inline policy references
# it, and the bootstrap step needs to read the scripts from it.
if ! aws s3api head-bucket --bucket "$ARTIFACT_BUCKET" --region "$REGION" >/dev/null 2>&1; then
log "creating artifact bucket ${ARTIFACT_BUCKET}"
if [[ "$REGION" == "us-east-1" ]]; then
aws s3api create-bucket --bucket "$ARTIFACT_BUCKET" --region "$REGION" >/dev/null
else
aws s3api create-bucket --bucket "$ARTIFACT_BUCKET" --region "$REGION" \
--create-bucket-configuration "LocationConstraint=${REGION}" >/dev/null
fi
aws s3api put-public-access-block --bucket "$ARTIFACT_BUCKET" \
--public-access-block-configuration \
"BlockPublicAcls=true,IgnorePublicAcls=true,BlockPublicPolicy=true,RestrictPublicBuckets=true" >/dev/null
aws s3api put-bucket-encryption --bucket "$ARTIFACT_BUCKET" \
--server-side-encryption-configuration \
'{"Rules":[{"ApplyServerSideEncryptionByDefault":{"SSEAlgorithm":"AES256"}}]}' >/dev/null
ok "bucket ready (public access blocked, SSE-S3 on)"
fi
aws s3 cp "$BOOTSTRAP_PS1" "s3://${ARTIFACT_BUCKET}/bootstrap.ps1" --region "$REGION" >/dev/null
aws s3 cp "$DIAGNOSE_PS1" "s3://${ARTIFACT_BUCKET}/diagnose.ps1" --region "$REGION" >/dev/null
ok "bootstrap.ps1 and diagnose.ps1 uploaded"
"${DEPLOY_CMD[@]}"
ok "stack ${STACK_NAME} deployed"
read -r INSTANCE_ID DCV_ENDPOINT ROLE_ARN <<<"$(aws cloudformation describe-stacks \
--stack-name "$STACK_NAME" --region "$REGION" \
--query 'Stacks[0].Outputs' --output text \
| awk '/InstanceId/{id=$2} /PublicDcvEndpoint/{ep=$2} /InstanceRoleArn/{rn=$2} END{print id, ep, rn}')"
log "instance ${INSTANCE_ID}"
log "dcv endpoint ${DCV_ENDPOINT}"
log "role arn ${ROLE_ARN}"
# ---------------------------------------------------------------- step 4: SSM
log "step 4/5 - waiting for the instance to register with Systems Manager"
DEADLINE=$(( SECONDS + 600 ))
while (( SECONDS < DEADLINE )); do
PING="$(aws ssm describe-instance-information \
--region "$REGION" \
--filters "Key=InstanceIds,Values=${INSTANCE_ID}" \
--query 'InstanceInformationList[0].PingStatus' --output text 2>/dev/null || echo None)"
if [[ "$PING" == "Online" ]]; then
ok "instance is Online in Systems Manager"
break
fi
printf ' ... %s (waiting)\n' "$PING"
sleep 20
done
[[ "$PING" == "Online" ]] || die "instance never registered with SSM; check the instance role and outbound 443"
# Register the diagnostic document so it is available from console, CLI or a schedule.
aws ssm create-document \
--name SteamGpu-Diagnose \
--document-type Command \
--document-format JSON \
--content "file://${SSM_DOC}" \
--region "$REGION" \
--tags Key=Project,Value=steam-gpu >/dev/null 2>&1 \
&& ok "SSM document SteamGpu-Diagnose registered" \
|| warn "SSM document already exists or could not be created; continuing"
# ---------------------------------------------------------------- step 5: bootstrap
if [[ "$BOOTSTRAP" == "1" ]]; then
log "step 5/5 - running the Windows bootstrap over SSM (10-20 minutes; the box reboots)"
CMD_ID="$(aws ssm send-command \
--region "$REGION" \
--instance-ids "$INSTANCE_ID" \
--document-name "AWS-RunPowerShellScript" \
--comment "steam-gpu bootstrap via launch.sh" \
--timeout-seconds 3600 \
--parameters "commands=[\"Set-ExecutionPolicy Bypass -Scope Process -Force; Read-S3Object -BucketName '${ARTIFACT_BUCKET}' -Key bootstrap.ps1 -File C:\\\\bootstrap.ps1; & C:\\\\bootstrap.ps1\"]" \
--query 'Command.CommandId' --output text)"
log "command id ${CMD_ID}"
log "follow along with:"
log " aws ssm list-command-invocations --command-id ${CMD_ID} --region ${REGION} --details \\"
log " --query 'CommandInvocations[0].CommandPlugins[0].Output' --output text"
else
log "step 5/5 - skipped (--no-bootstrap)"
warn "upload the DCV MSI and driver yourself, then run scripts/bootstrap.ps1 from the instance"
fi
printf '\n'
ok "launch complete"
cat <<EOF
Next steps
1. Upload the Amazon DCV server MSI to the instance (C:\\DCV) and re-run the DCV
section of scripts/bootstrap.ps1, then reboot for the GPU driver.
2. Confirm the licence:
nvidia-smi -q
expect: Product Name : NVIDIA Cloud Gaming / License Status : Licensed
3. Open ${DCV_ENDPOINT} in a browser and connect to the console session.
4. Configure Steam Offline Mode (see README): save credentials, clear the queue,
launch each game once online, then Steam > Go Offline.
5. Run the health report:
aws ssm send-command --document-name SteamGpu-Diagnose \\
--instance-ids ${INSTANCE_ID} --region ${REGION}
RDP without exposing 3389
aws ssm start-session --target ${INSTANCE_ID} --region ${REGION} \\
--document-name AWS-StartPortForwardingSession \\
--parameters '{"portNumber":["3389"],"localPortNumber":["13389"]}'
Cost
${INSTANCE_TYPE} runs until you stop it. Stop ${INSTANCE_ID} between sessions.
EOF