Merge pull request #575 from buildkite/remove-aws-autoscaling

Remove AWS Autoscaling in favor of buildkite-agent-scaler

Author: Lachlan Donald

.buildkite/steps/launch.sh

@@ -84,10 +84,6 @@ cat << EOF > config.json
   {
     "ParameterKey": "EnableAgentGitMirrorsExperiment",
     "ParameterValue": "true"
-  },
-  {
-    "ParameterKey": "EnableExperimentalLambdaBasedAutoscaling",
-    "ParameterValue": "true"
   }
 ]
 EOF

packer/linux/conf/bin/bk-install-elastic-stack.sh

@@ -116,34 +116,32 @@ spawn=${BUILDKITE_AGENTS_PER_INSTANCE}
 no-color=true
 EOF
 
+terminate_instance_on_agent_stop() {
+  mkdir -p /etc/systemd/system/buildkite-agent.service.d/
+  cat << EOF > /etc/systemd/system/buildkite-agent.service.d/10-power-off-stop.conf
+[Service]
+ExecStopPost=/usr/local/bin/terminate-instance
+ExecStopPost=/bin/sudo poweroff
+EOF
+
+  # If we modify the systemd, we need to rebuild the dependency tree
+  systemctl daemon-reload
+}
+
 # Add conditional config to the agent config
 if [[ "${BUILDKITE_TERMINATE_INSTANCE_AFTER_JOB:-false}" == "true" ]] ; then
   cat << EOF >> /etc/buildkite-agent/buildkite-agent.cfg
 disconnect-after-job=true
 disconnect-after-job-timeout=${BUILDKITE_TERMINATE_INSTANCE_AFTER_JOB_TIMEOUT}
 EOF
-elif [[ "${BUILDKITE_LAMBDA_AUTOSCALING}" == "true" ]] ; then
-  cat << EOF >> /etc/buildkite-agent/buildkite-agent.cfg
-disconnect-after-idle-timeout=${BUILDKITE_SCALE_DOWN_PERIOD}
-EOF
+  terminate_instance_on_agent_stop
 fi
 
-# If we are using the lambda for autoscaling, always decrease capacity
 if [[ "${BUILDKITE_LAMBDA_AUTOSCALING}" == "true" ]] ; then
-  BUILDKITE_TERMINATE_INSTANCE_AFTER_JOB_DECREASE_DESIRED_CAPACITY=true
-fi
-
-if [[ "${BUILDKITE_TERMINATE_INSTANCE_AFTER_JOB}" == "true" || "${BUILDKITE_LAMBDA_AUTOSCALING}" == "true" ]] ; then
-  mkdir -p /etc/systemd/system/buildkite-agent.service.d/
-
-  cat << EOF > /etc/systemd/system/buildkite-agent.service.d/10-power-off-stop.conf
-[Service]
-ExecStopPost=/usr/local/bin/terminate-instance ${BUILDKITE_TERMINATE_INSTANCE_AFTER_JOB_DECREASE_DESIRED_CAPACITY}
-ExecStopPost=/bin/sudo poweroff
+  cat << EOF >> /etc/buildkite-agent/buildkite-agent.cfg
+disconnect-after-idle-timeout=${BUILDKITE_SCALE_IN_IDLE_PERIOD}
 EOF
-
-  # If we modify the systemd, we need to rebuild the dependency tree
-  systemctl daemon-reload
+  terminate_instance_on_agent_stop
 fi
 
 chown buildkite-agent: /etc/buildkite-agent/buildkite-agent.cfg
@@ -168,7 +166,6 @@ fi
 
 cat << EOF > /etc/lifecycled
 AWS_REGION=${AWS_REGION}
-LIFECYCLED_SNS_TOPIC=${BUILDKITE_LIFECYCLE_TOPIC}
 LIFECYCLED_HANDLER=/usr/local/bin/stop-agent-gracefully
 LIFECYCLED_CLOUDWATCH_GROUP=/buildkite/lifecycled
 EOF

packer/windows/conf/bin/bk-install-elastic-stack.ps1

@@ -99,15 +99,11 @@ disconnect-after-job=true
 disconnect-after-job-timeout=$Env:BUILDKITE_TERMINATE_INSTANCE_AFTER_JOB_TIMEOUT
 "@
 }
-ElseIf ($Env:BUILDKITE_LAMBDA_AUTOSCALING -eq "true") {
-  Add-Content -Path C:\buildkite-agent\buildkite-agent.cfg -Value @"
-disconnect-after-idle-timeout=${Env:BUILDKITE_SCALE_DOWN_PERIOD}
-"@
-}
 
-# If we are using the lambda for autoscaling, always decrease capacity
 If ($Env:BUILDKITE_LAMBDA_AUTOSCALING -eq "true") {
-  $Env:BUILDKITE_TERMINATE_INSTANCE_AFTER_JOB_DECREASE_DESIRED_CAPACITY = "true"
+  Add-Content -Path C:\buildkite-agent\buildkite-agent.cfg -Value @"
+disconnect-after-idle-timeout=${Env:BUILDKITE_SCALE_IN_IDLE_PERIOD}
+"@
 }
 
 If (![string]::IsNullOrEmpty($Env:BUILDKITE_ELASTIC_BOOTSTRAP_SCRIPT)) {
@@ -117,7 +113,6 @@ If (![string]::IsNullOrEmpty($Env:BUILDKITE_ELASTIC_BOOTSTRAP_SCRIPT)) {
 }
 
 nssm set lifecycled AppEnvironmentExtra :AWS_REGION=$Env:AWS_REGION
-nssm set lifecycled AppEnvironmentExtra +LIFECYCLED_SNS_TOPIC=$Env:BUILDKITE_LIFECYCLE_TOPIC
 nssm set lifecycled AppEnvironmentExtra +LIFECYCLED_HANDLER="C:\buildkite-agent\bin\stop-agent-gracefully.ps1"
 Restart-Service lifecycled
 
@@ -155,9 +150,10 @@ nssm set buildkite-agent ObjectName .\$UserName $Password
 nssm set buildkite-agent AppStdout C:\buildkite-agent\buildkite-agent.log
 nssm set buildkite-agent AppStderr C:\buildkite-agent\buildkite-agent.log
 nssm set buildkite-agent AppEnvironmentExtra :HOME=C:\buildkite-agent
+
 If (($Env:BUILDKITE_TERMINATE_INSTANCE_AFTER_JOB -eq "true") -or ($Env:BUILDKITE_LAMBDA_AUTOSCALING -eq "true")) {
   nssm set buildkite-agent AppExit Default Exit
-  nssm set buildkite-agent AppEvents Exit/Post "powershell C:\buildkite-agent\bin\terminate-instance.ps1 $Env:BUILDKITE_TERMINATE_INSTANCE_AFTER_JOB_DECREASE_DESIRED_CAPACITY"
+  nssm set buildkite-agent AppEvents Exit/Post "powershell C:\buildkite-agent\bin\terminate-instance.ps1"
 }
 Restart-Service buildkite-agent