Replace lifecycled and self-termination with ASG lifecycle hooks

goss.yaml

@@ -44,10 +44,6 @@ service:
     enabled: true
     running: true
 
-  lifecycled:
-    enabled: true
-    running: true
-
   sshd:
     enabled: true
     running: true
@@ -93,9 +89,6 @@ process:
   buildkite-agent:
     running: true
 
-  lifecycled:
-    running: true
-
   sshd:
     running: true
 

packer/linux/buildkite-ami.json

@@ -45,10 +45,6 @@
       "type": "shell",
       "script": "scripts/install-cloudwatch-agent.sh"
     },
-    {
-      "type": "shell",
-      "script": "scripts/install-lifecycled.sh"
-    },
     {
       "type": "shell",
       "script": "scripts/install-docker.sh"

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

@@ -176,7 +176,6 @@ experiment="${BUILDKITE_AGENT_EXPERIMENTS}"
 priority=%n
 spawn=${BUILDKITE_AGENTS_PER_INSTANCE}
 no-color=true
-disconnect-after-idle-timeout=${BUILDKITE_SCALE_IN_IDLE_PERIOD}
 disconnect-after-job=${BUILDKITE_TERMINATE_INSTANCE_AFTER_JOB}
 EOF
 
@@ -203,15 +202,6 @@ if [[ -n "${BUILDKITE_ELASTIC_BOOTSTRAP_SCRIPT}" ]] ; then
 	rm /tmp/elastic_bootstrap
 fi
 
-cat << EOF > /etc/lifecycled
-AWS_REGION=${AWS_REGION}
-LIFECYCLED_HANDLER=/usr/local/bin/stop-agent-gracefully
-LIFECYCLED_CLOUDWATCH_GROUP=/buildkite/lifecycled
-EOF
-
-systemctl enable lifecycled.service
-systemctl start lifecycled
-
 # wait for docker to start
 next_wait_time=0
 until docker ps || [ $next_wait_time -eq 5 ]; do
@@ -224,7 +214,6 @@ if ! docker ps ; then
 fi
 
 systemctl enable "buildkite-agent"
-systemctl start "buildkite-agent"
 
 # let the stack know that this host has been initialized successfully
 /opt/aws/bin/cfn-signal \

packer/linux/conf/buildkite-agent/systemd/buildkite-agent.service

@@ -12,9 +12,8 @@ Environment="HOME=/var/lib/buildkite-agent"
 Environment="PATH=/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin"
 Environment="USER=buildkite-agent"
 ExecStart=/usr/bin/buildkite-agent start
-ExecStopPost=/usr/local/bin/terminate-instance
 RestartSec=5
-Restart=on-failure
+Restart=always
 RestartForceExitStatus=SIGPIPE
 TimeoutStartSec=10
 TimeoutStopSec=0

packer/windows/buildkite-ami.json

@@ -44,10 +44,6 @@
       "type": "powershell",
       "script": "scripts/install-cloudwatch-agent.ps1"
     },
-    {
-      "type": "powershell",
-      "script": "scripts/install-lifecycled.ps1"
-    },
     {
       "type": "powershell",
       "script": "scripts/install-docker.ps1"

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

@@ -135,15 +135,10 @@ priority=%n
 spawn=${Env:BUILDKITE_AGENTS_PER_INSTANCE}
 no-color=true
 shell=powershell
-disconnect-after-idle-timeout=${Env:BUILDKITE_SCALE_IN_IDLE_PERIOD}
 disconnect-after-job=${Env:BUILDKITE_TERMINATE_INSTANCE_AFTER_JOB}
 "@
 $OFS=" "
 
-nssm set lifecycled AppEnvironmentExtra :AWS_REGION=$Env:AWS_REGION
-nssm set lifecycled AppEnvironmentExtra +LIFECYCLED_HANDLER="C:\buildkite-agent\bin\stop-agent-gracefully.ps1"
-Restart-Service lifecycled
-
 # wait for docker to start
 $next_wait_time=0
 do {
@@ -212,13 +207,9 @@ nssm set buildkite-agent AppEnvironmentExtra :HOME=C:\buildkite-agent
 If ($lastexitcode -ne 0) { Exit $lastexitcode }
 nssm set buildkite-agent AppExit Default Restart
 If ($lastexitcode -ne 0) { Exit $lastexitcode }
-nssm set buildkite-agent AppRestartDelay 10000
-If ($lastexitcode -ne 0) { Exit $lastexitcode }
-nssm set buildkite-agent AppEvents Exit/Post "powershell C:\buildkite-agent\bin\terminate-instance.ps1"
+nssm set buildkite-agent AppRestartDelay 5000
 If ($lastexitcode -ne 0) { Exit $lastexitcode }
 
-Restart-Service buildkite-agent
-
 # renable debug tracing
 Set-PSDebug -Trace 2