AWS Security ChangesHomeSearch

AWS parallelcluster: ParallelCluster v3 IAM roles: login nodes, secrets, and scoped permissions

Service: parallelcluster · 2026-10-23 · Documentation medium

File: parallelcluster/latest/ug/iam-roles-in-parallelcluster-v3.md · Type: iam

Summary

Updates the AWS ParallelCluster v3 IAM roles documentation: adds a new 'Additional pcluster user policy when using login nodes' section, a new 'Login nodes IAM role' section with a minimal policy, adds secretsmanager:GetSecretValue to the secrets-validation statement, adds resource-groups/tag/cloudformation/ec2/elb permissions, and adds guidance to scope down secrets permissions to only the secrets in the cluster configuration.

Security assessment

The diff documents IAM policies and best practices for ParallelCluster v3, including least-privilege guidance to scope secretsmanager permissions to only the secrets referenced in the cluster configuration, and adds a new login-nodes IAM role policy. This is security hardening documentation (IAM permissions, secrets access) rather than a fix for a specific vulnerability or incident, so security_issue_related is false while adds_security_documentation is true.

Evidence

+To improve your security posture, scope down these permissions to only the secrets that you specify in your cluster configuration.

Diff

diff --git a/parallelcluster/latest/ug/iam-roles-in-parallelcluster-v3.md b/parallelcluster/latest/ug/iam-roles-in-parallelcluster-v3.md
index 66c050f..de81568 100644
--- a//parallelcluster/latest/ug/iam-roles-in-parallelcluster-v3.md
+++ b//parallelcluster/latest/ug/iam-roles-in-parallelcluster-v3.md
@@ -68,0 +69,2 @@ The following examples show the user policies required to create and manage AWS
+  * Additional AWS ParallelCluster pcluster user policy when using login nodes
+
@@ -80 +82,3 @@ The following policy shows the permissions required to run AWS ParallelCluster `
-The last action listed in the policy is included to provide validation of any secrets specified in the cluster configuration. For example, an AWS Secrets Manager secret is used to configure the [DirectoryService](./DirectoryService-v3.html) integration. In this case, a cluster is created only if a valid secret exists in the [PasswordSecretArn](./DirectoryService-v3.html#yaml-DirectoryService-PasswordSecretArn). If this action is omitted, secret validation is skipped. To improve your security posture, we recommend that you scope down this policy statement by adding only the secrets specified in your cluster configuration.
+To have AWS ParallelCluster validate the secrets that you specify in your cluster configuration, such as [DirectoryService / PasswordSecretArn](./DirectoryService-v3.html#yaml-DirectoryService-PasswordSecretArn) and [MungeKeySecretArn](./Scheduling-v3.html#yaml-Scheduling-SlurmSettings-MungeKeySecretArn), before it creates the cluster, grant the `secretsmanager:DescribeSecret` and `secretsmanager:GetSecretValue` permissions. If you omit those permissions, the validation is skipped.
+
+To improve your security posture, scope down these permissions to only the secrets that you specify in your cluster configuration.
@@ -322 +326,2 @@ JSON
-                    "resource-groups:ListGroupResources"
+                    "resource-groups:ListGroupResources",
+                    "resource-groups:GetGroupConfiguration"
@@ -327,0 +333,8 @@ JSON
+            {
+                "Action": [
+                    "tag:GetResources"
+                ],
+                "Resource": "*",
+                "Effect": "Allow",
+                "Sid": "ResourceGroupsTaggingApiRead"
+            },
@@ -337 +350,4 @@ JSON
-                "Action": "secretsmanager:DescribeSecret",
+                "Action": [
+                    "secretsmanager:DescribeSecret",
+                    "secretsmanager:GetSecretValue"
+                ],
@@ -346,0 +363,4 @@ JSON
+###### Note
+
+AWS ParallelCluster version 3.15.1 and earlier supports AWS Batch as a scheduler.
+
@@ -438,0 +459,86 @@ JSON
+### Additional AWS ParallelCluster `pcluster` user policy when using login nodes
+
+To create and manage a cluster with [LoginNodes](./LoginNodes-v3.html), add the following policy to the `pcluster` user. This policy grants the permissions to create the required service-linked roles and to manage the Auto Scaling groups and load balancers that the login nodes use:
+    
+    
+    {
+        "Version": "2012-10-17",
+        "Statement": [
+            {
+                "Action": [
+                    "iam:CreateServiceLinkedRole",
+                    "iam:DeleteServiceLinkedRole"
+                ],
+                "Resource": "*",
+                "Effect": "Allow",
+                "Condition": {
+                    "StringEquals": {
+                        "iam:AWSServiceName": [
+                            "elasticloadbalancing.amazonaws.com",
+                            "autoscaling.amazonaws.com"
+                        ]
+                    }
+                },
+                "Sid": "LoginNodesServiceLinkedRoles"
+            },
+            {
+                "Action": [
+                    "autoscaling:DeleteAutoScalingGroup",
+                    "autoscaling:DeleteLifecycleHook",
+                    "autoscaling:DescribeAutoScalingGroups",
+                    "autoscaling:DescribeLifecycleHooks",
+                    "autoscaling:DescribeScalingActivities",
+                    "autoscaling:PutLifecycleHook",
+                    "autoscaling:UpdateAutoScalingGroup",
+                    "elasticloadbalancing:DescribeTags",
+                    "elasticloadbalancing:DeleteListener",
+                    "elasticloadbalancing:DeleteLoadBalancer",
+                    "elasticloadbalancing:DeleteTargetGroup",
+                    "elasticloadbalancing:DescribeListeners",
+                    "elasticloadbalancing:DescribeLoadBalancers",
+                    "elasticloadbalancing:DescribeTargetGroups",
+                    "elasticloadbalancing:DescribeTargetHealth",
+                    "elasticloadbalancing:ModifyLoadBalancerAttributes"
+                ],
+                "Resource": "*",
+                "Effect": "Allow",
+                "Condition": {
+                    "ForAllValues:StringEquals": {
+                        "aws:TagKeys": [
+                            "parallelcluster:cluster-name"
+                        ]
+                    }
+                },
+                "Sid": "LoginNodesFunctionalities"
+            },
+            {
+                "Action": [
+                    "elasticloadbalancing:CreateListener",
+                    "elasticloadbalancing:CreateTargetGroup"
+                ],
+                "Resource": "*",
+                "Effect": "Allow",
+                "Condition": {
+                    "Null": {
+                        "aws:RequestTag/parallelcluster:cluster-name": "false"
+                    }
+                },
+                "Sid": "LoginNodesFunctionalitiesCondition2"
+            },
+            {
+                "Action": [
+                    "autoscaling:CreateAutoScalingGroup",
+                    "autoscaling:DeleteTags",
+                    "autoscaling:CreateOrUpdateTags",
+                    "elasticloadbalancing:AddTags",
+                    "elasticloadbalancing:RemoveTags",
+                    "elasticloadbalancing:ModifyTargetGroup",
+                    "elasticloadbalancing:CreateLoadBalancer"
+                ],
+                "Resource": "*",
+                "Effect": "Allow",
+                "Sid": "LoginNodesFunctionalitiesNoCondition"
+            }
+        ]
+    }
+
@@ -570,0 +677,18 @@ JSON
+            {
+                "Condition": {
+                    "ArnLike": {
+                        "iam:PolicyARN": [
+                            "arn:aws:iam::aws:policy/service-role/AWSLambdaBasicExecutionRole",
+                            "arn:aws:iam::aws:policy/service-role/AWSLambdaVPCAccessExecutionRole"
+                        ]
+                    }
+                },
+                "Action": [
+                    "iam:AttachRolePolicy"
+                ],
+                "Resource": [
+                    "arn:aws:iam::111122223333:role/parallelcluster/*"
+                ],
+                "Effect": "Allow",
+                "Sid": "IAMCleanupRole"
+            },
@@ -597,0 +722,8 @@ JSON
+            {
+                "Action": [
+                    "cloudformation:DescribeStacks"
+                ],
+                "Resource": "*",
+                "Effect": "Allow",
+                "Sid": "CloudFormationListImages"
+            },
@@ -631,0 +764 @@ JSON
+                    "imagebuilder:CancelImageCreation",
@@ -821,0 +955,7 @@ When creating a cluster the following parameters are required:
+When creating a cluster with login nodes, the following parameter is also required:
+
+  * [LoginNodes](./LoginNodes-v3.html) / [Pools](./LoginNodes-v3.html#LoginNodes-v3-Pools) / [Iam](./LoginNodes-v3.html#yaml-LoginNodes-Pools-Iam) / [InstanceRole](./LoginNodes-v3.html#yaml-LoginNodes-Pools-Iam-InstanceRole) | [InstanceProfile](./LoginNodes-v3.html#yaml-LoginNodes-Pools-Iam-InstanceProfile)
+
+
+
+
@@ -887 +1027,2 @@ JSON
-                    "iam:TagRole"
+                    "iam:TagRole",
+                    "iam:UntagRole"
@@ -992,0 +1134,2 @@ AWS ParallelCluster exposes a series of configuration options to customize and m
+  * Login nodes IAM role
+
@@ -1032,0 +1176,10 @@ JSON
+            {
+                "Action": [
+                    "s3:GetBucketLocation",
+                    "s3:ListBucket"
+                ],
+                "Resource": [
+                    "arn:aws:s3:::parallelcluster-*-v1-do-not-delete"
+                ],
+                "Effect": "Allow"
+            },
@@ -1080,0 +1234 @@ JSON
+                    "ec2:DescribeLaunchTemplateVersions",
@@ -1093 +1247,2 @@ JSON
-                    "ec2:AttachVolume"
+                    "ec2:AttachVolume",
+                    "ec2:DetachVolume"
@@ -1100,0 +1256,18 @@ JSON
+            {
+                "Condition": {
+                    "StringEquals": {
+                        "aws:ResourceTag/parallelcluster:node-type": "Compute"
+                    }
+                },
+                "Action": "ec2:GetConsoleOutput",
+                "Resource": "arn:aws:ec2:us-east-1:111122223333:instance/*",
+                "Effect": "Allow"
+            },
+            {
+                "Action": [
+                    "elasticloadbalancing:DescribeTargetGroups",
+                    "elasticloadbalancing:DescribeTargetHealth"