fix: Windows EC2 runner SSM agent stopped working (#567)

I am not sure when this started happening, but EC2Launch won't start SSM Agent until user data finishes executing. Since our user data runs until the very end, that never happens. Fix this by manually starting up SSM Agent.

Fixes #564

src/providers/ec2.ts

@@ -113,6 +113,10 @@ $runnerTokenPath="{}"
 $labels="{}"
 $registrationURL="{}"
 
+# EC2Launch only starts ssm agent after user data is done, so we need to start it ourselves (it is disabled by default)
+Set-Service -StartupType Manual AmazonSSMAgent
+Start-Service AmazonSSMAgent
+
 Start-Job -ScriptBlock {
   while (1) {
     aws stepfunctions send-task-heartbeat --task-token "$using:TASK_TOKEN"

test/default.integ.snapshot/github-runners-test.assets.json

@@ -209,15 +209,15 @@
         }
       }
     },
-    "7e31d5548fbce527f8c0e4c0bab79dc8ae721e24c4f2934f30d81301929ae02d": {
+    "c7adc373a2e617c07b79256b0990e1a6ed3b10843f562ffcc8a783769346dce1": {
       "source": {
         "path": "github-runners-test.template.json",
         "packaging": "file"
       },
       "destinations": {
         "current_account-current_region": {
           "bucketName": "cdk-hnb659fds-assets-${AWS::AccountId}-${AWS::Region}",
-          "objectKey": "7e31d5548fbce527f8c0e4c0bab79dc8ae721e24c4f2934f30d81301929ae02d.json",
+          "objectKey": "c7adc373a2e617c07b79256b0990e1a6ed3b10843f562ffcc8a783769346dce1.json",
           "assumeRoleArn": "arn:${AWS::Partition}:iam::${AWS::AccountId}:role/cdk-hnb659fds-file-publishing-role-${AWS::AccountId}-${AWS::Region}"
         }
       }

test/default.integ.snapshot/github-runners-test.template.json

@@ -16123,7 +16123,7 @@
        {
         "Ref": "EC2Linuxarm64AMIRootDevice3046D37D"
        },
-       "\",\"Ebs\":{\"DeleteOnTermination\":true,\"VolumeSize\":30}}]}},\"ec2, windows, x64 data\":{\"Type\":\"Pass\",\"ResultPath\":\"$.ec2\",\"Parameters\":{\"userdataTemplate\":\"<powershell>\\n$TASK_TOKEN = \\\"{}\\\"\\n$logGroupName=\\\"{}\\\"\\n$runnerNamePath=\\\"{}\\\"\\n$githubDomainPath=\\\"{}\\\"\\n$ownerPath=\\\"{}\\\"\\n$repoPath=\\\"{}\\\"\\n$runnerTokenPath=\\\"{}\\\"\\n$labels=\\\"{}\\\"\\n$registrationURL=\\\"{}\\\"\\n\\nStart-Job -ScriptBlock \\\\{\\n  while (1) \\\\{\\n    aws stepfunctions send-task-heartbeat --task-token \\\"$using:TASK_TOKEN\\\"\\n    sleep 60\\n  \\\\}\\n\\\\}\\nfunction setup_logs () \\\\{\\n  echo \\\"\\\\{\\n    `\\\"logs`\\\": \\\\{\\n      `\\\"log_stream_name`\\\": `\\\"unknown`\\\",\\n      `\\\"logs_collected`\\\": \\\\{\\n        `\\\"files`\\\": \\\\{\\n         `\\\"collect_list`\\\": [\\n            \\\\{\\n              `\\\"file_path`\\\": `\\\"/actions/runner.log`\\\",\\n              `\\\"log_group_name`\\\": `\\\"$logGroupName`\\\",\\n              `\\\"log_stream_name`\\\": `\\\"$runnerNamePath`\\\",\\n              `\\\"timezone`\\\": `\\\"UTC`\\\"\\n            \\\\}\\n          ]\\n        \\\\}\\n      \\\\}\\n    \\\\}\\n  \\\\}\\\" | Out-File -Encoding ASCII $Env:TEMP/log.conf\\n  & \\\"C:/Program Files/Amazon/AmazonCloudWatchAgent/amazon-cloudwatch-agent-ctl.ps1\\\" -a fetch-config -m ec2 -s -c file:$Env:TEMP/log.conf\\n\\\\}\\nfunction action () \\\\{\\n  cd /actions\\n  $RunnerVersion = Get-Content RUNNER_VERSION -Raw\\n  if ($RunnerVersion -eq \\\"latest\\\") \\\\{ $RunnerFlags = \\\"\\\" \\\\} else \\\\{ $RunnerFlags = \\\"--disableupdate\\\" \\\\}\\n  ./config.cmd --unattended --url \\\"$\\\\{registrationUrl\\\\}\\\" --token \\\"$\\\\{runnerTokenPath\\\\}\\\" --ephemeral --work _work --labels \\\"$\\\\{labels\\\\},cdkghr:started:$(Get-Date -UFormat +%s)\\\" $RunnerFlags --name \\\"$\\\\{runnerNamePath\\\\}\\\" 2>&1 | Out-File -Encoding ASCII -Append /actions/runner.log\\n\\n  if ($LASTEXITCODE -ne 0) \\\\{ return 1 \\\\}\\n  ./run.cmd 2>&1 | Out-File -Encoding ASCII -Append /actions/runner.log\\n  if ($LASTEXITCODE -ne 0) \\\\{ return 2 \\\\}\\n\\n  $STATUS = Select-String -Path './_diag/*.log' -Pattern 'finish job request for job [0-9a-f\\\\-]+ with result: (.*)' | %\\\\{$_.Matches.Groups[1].Value\\\\} | Select-Object -Last 1\\n\\n  if ($STATUS) \\\\{\\n      echo \\\"CDKGHA JOB DONE $\\\\{labels\\\\} $STATUS\\\" | Out-File -Encoding ASCII -Append /actions/runner.log\\n  \\\\}\\n\\n  return 0\\n\\\\}\\nsetup_logs\\n$r = action\\nif ($r -eq 0) \\\\{\\n  aws stepfunctions send-task-success --task-token \\\"$TASK_TOKEN\\\" --task-output '\\\\{ \\\\}'\\n\\\\} else \\\\{\\n  aws stepfunctions send-task-failure --task-token \\\"$TASK_TOKEN\\\"\\n\\\\}\\nStart-Sleep -Seconds 10  # give cloudwatch agent its default 5 seconds buffer duration to upload logs\\nStop-Computer -ComputerName localhost -Force\\n</powershell>\\n\"},\"Next\":\"ec2, windows, x64 subnet1\"},\"ec2, windows, x64 subnet1\":{\"End\":true,\"Catch\":[{\"ErrorEquals\":[\"Ec2.Ec2Exception\",\"States.Timeout\"],\"ResultPath\":\"$.lastSubnetError\",\"Next\":\"ec2, windows, x64 subnet2\"}],\"Type\":\"Task\",\"Comment\":\"",
+       "\",\"Ebs\":{\"DeleteOnTermination\":true,\"VolumeSize\":30}}]}},\"ec2, windows, x64 data\":{\"Type\":\"Pass\",\"ResultPath\":\"$.ec2\",\"Parameters\":{\"userdataTemplate\":\"<powershell>\\n$TASK_TOKEN = \\\"{}\\\"\\n$logGroupName=\\\"{}\\\"\\n$runnerNamePath=\\\"{}\\\"\\n$githubDomainPath=\\\"{}\\\"\\n$ownerPath=\\\"{}\\\"\\n$repoPath=\\\"{}\\\"\\n$runnerTokenPath=\\\"{}\\\"\\n$labels=\\\"{}\\\"\\n$registrationURL=\\\"{}\\\"\\n\\n# EC2Launch only starts ssm agent after user data is done, so we need to start it ourselves (it is disabled by default)\\nSet-Service -StartupType Manual AmazonSSMAgent\\nStart-Service AmazonSSMAgent\\n\\nStart-Job -ScriptBlock \\\\{\\n  while (1) \\\\{\\n    aws stepfunctions send-task-heartbeat --task-token \\\"$using:TASK_TOKEN\\\"\\n    sleep 60\\n  \\\\}\\n\\\\}\\nfunction setup_logs () \\\\{\\n  echo \\\"\\\\{\\n    `\\\"logs`\\\": \\\\{\\n      `\\\"log_stream_name`\\\": `\\\"unknown`\\\",\\n      `\\\"logs_collected`\\\": \\\\{\\n        `\\\"files`\\\": \\\\{\\n         `\\\"collect_list`\\\": [\\n            \\\\{\\n              `\\\"file_path`\\\": `\\\"/actions/runner.log`\\\",\\n              `\\\"log_group_name`\\\": `\\\"$logGroupName`\\\",\\n              `\\\"log_stream_name`\\\": `\\\"$runnerNamePath`\\\",\\n              `\\\"timezone`\\\": `\\\"UTC`\\\"\\n            \\\\}\\n          ]\\n        \\\\}\\n      \\\\}\\n    \\\\}\\n  \\\\}\\\" | Out-File -Encoding ASCII $Env:TEMP/log.conf\\n  & \\\"C:/Program Files/Amazon/AmazonCloudWatchAgent/amazon-cloudwatch-agent-ctl.ps1\\\" -a fetch-config -m ec2 -s -c file:$Env:TEMP/log.conf\\n\\\\}\\nfunction action () \\\\{\\n  cd /actions\\n  $RunnerVersion = Get-Content RUNNER_VERSION -Raw\\n  if ($RunnerVersion -eq \\\"latest\\\") \\\\{ $RunnerFlags = \\\"\\\" \\\\} else \\\\{ $RunnerFlags = \\\"--disableupdate\\\" \\\\}\\n  ./config.cmd --unattended --url \\\"$\\\\{registrationUrl\\\\}\\\" --token \\\"$\\\\{runnerTokenPath\\\\}\\\" --ephemeral --work _work --labels \\\"$\\\\{labels\\\\},cdkghr:started:$(Get-Date -UFormat +%s)\\\" $RunnerFlags --name \\\"$\\\\{runnerNamePath\\\\}\\\" 2>&1 | Out-File -Encoding ASCII -Append /actions/runner.log\\n\\n  if ($LASTEXITCODE -ne 0) \\\\{ return 1 \\\\}\\n  ./run.cmd 2>&1 | Out-File -Encoding ASCII -Append /actions/runner.log\\n  if ($LASTEXITCODE -ne 0) \\\\{ return 2 \\\\}\\n\\n  $STATUS = Select-String -Path './_diag/*.log' -Pattern 'finish job request for job [0-9a-f\\\\-]+ with result: (.*)' | %\\\\{$_.Matches.Groups[1].Value\\\\} | Select-Object -Last 1\\n\\n  if ($STATUS) \\\\{\\n      echo \\\"CDKGHA JOB DONE $\\\\{labels\\\\} $STATUS\\\" | Out-File -Encoding ASCII -Append /actions/runner.log\\n  \\\\}\\n\\n  return 0\\n\\\\}\\nsetup_logs\\n$r = action\\nif ($r -eq 0) \\\\{\\n  aws stepfunctions send-task-success --task-token \\\"$TASK_TOKEN\\\" --task-output '\\\\{ \\\\}'\\n\\\\} else \\\\{\\n  aws stepfunctions send-task-failure --task-token \\\"$TASK_TOKEN\\\"\\n\\\\}\\nStart-Sleep -Seconds 10  # give cloudwatch agent its default 5 seconds buffer duration to upload logs\\nStop-Computer -ComputerName localhost -Force\\n</powershell>\\n\"},\"Next\":\"ec2, windows, x64 subnet1\"},\"ec2, windows, x64 subnet1\":{\"End\":true,\"Catch\":[{\"ErrorEquals\":[\"Ec2.Ec2Exception\",\"States.Timeout\"],\"ResultPath\":\"$.lastSubnetError\",\"Next\":\"ec2, windows, x64 subnet2\"}],\"Type\":\"Task\",\"Comment\":\"",
        {
         "Ref": "VpcPublicSubnet1Subnet5C2D37C4"
        },