Troubleshooting Session Manager

Use the following information to help you troubleshoot problems with AWS Systems Manager Session Manager.

AccessDeniedException when calling the TerminateSession operation

Problem: When attempting to terminate a session, Systems Manager returns the following error:

An error occurred (AccessDeniedException) when calling the TerminateSession operation: 
User: <user_arn> is not authorized to perform: ssm:TerminateSession on resource: 
<ssm_session_arn> because no identity-based policy allows the ssm:TerminateSession action.

Solution A: Confirm that the latest version of the Session Manager plugin is installed on the node

Enter the following command in the terminal and press Enter.

session-manager-plugin --version

Solution B: Install or reinstall the latest version of the plugin

For more information, see Install the Session Manager plugin for the AWS CLI.

Solution C: Attempt to reestablish a connection to the node

Verify that the node is responding to requests. Try reestablishing the session. Or, if necessary, open the Amazon EC2 console and verify the status of the instance is running.

Document process failed unexpectedly: document worker timed out

Problem: When starting a session to a Linux host, Systems Manager returns the following error:

document process failed unexpectedly: document worker timed out, 
check [ssm-document-worker]/[ssm-session-worker] log for crash reason

If you configured SSM Agent logging, as described in Viewing SSM Agent logs, you can view more details in the debugging log. For this issue, Session Manager shows the following log entry:

failed to create channel: too many open files

This error typically indicates that there are too many Session Manager worker processes running and the underlying operating system reached a limit. You have two options for resolving this issue.

Solution A: Increase the operating system file notification limit

You can increase the limit by running the following command from a separate Linux host. This command uses Systems Manager Run Command. The specified value increases max_user_instances to 8192. This value is considerably higher than the default value of 128, but it won't strain host resources:

aws ssm send-command --document-name AWS-RunShellScript \
--instance-id i-02573cafcfEXAMPLE  --parameters \
"commands=sudo sysctl fs.inotify.max_user_instances=8192"

Solution B: Decrease the file notifications used by Session Manager in the target host

Run the following command from a separate Linux host to list sessions running on the target host:

aws ssm describe-sessions --state Active --filters key=Target,value=i-02573cafcfEXAMPLE

Review the command output to identify sessions that are no longer needed. You can terminate those session by running the following command from a separate Linux host:

aws ssm terminate-session —session-id session ID

Optionally, once there are no more sessions running on the remote server, you can free additional resources by running the following command from a separate Linux host. This command terminates all Session Manager processes running on the remote host, and consequently all sessions to the remote host. Before you run this command, verify there are no ongoing sessions you would like to keep:

aws ssm send-command --document-name AWS-RunShellScript \
            --instance-id i-02573cafcfEXAMPLE --parameters \
'{"commands":["sudo kill $(ps aux | grep ssm-session-worker | grep -v grep | awk '"'"'{print $2}'"'"')"]}'

Session Manager can't connect from the Amazon EC2 console

Problem: After creating a new instance, the Connect button > Session Manager tab in the Amazon Elastic Compute Cloud (Amazon EC2) console doesn't give you the option to connect.

Solution A: Create an instance profile: If you haven't already done so (as instructed by the information on the Session Manager tab in the EC2 console), create an AWS Identity and Access Management (IAM) instance profile by using Quick Setup. Quick Setup is a tool in AWS Systems Manager.

Session Manager requires an IAM instance profile to connect to your instance. You can create an instance profile and assign it to your instance by creating a host management configuration with Quick Setup. A host management configuration creates an instance profile with the required permissions and assigns it to your instance. A host management configuration also enables other Systems Manager tools and creates IAM roles for running those tools. There is no charge to use Quick Setup or the tools enabled by the host management configuration. Open Quick Setup and create a host management configuration.

Important

After you create the host management configuration, Amazon EC2 can take several minutes to register the change and refresh the Session Manager tab. If the tab doesn't show you a Connect button after two minutes, reboot your instance. After it reboots, if you still don't see the option to connect, open Quick Setup and verify you have only one host management configuration. If there are two, delete the older configuration and wait a few minutes.

If you still can't connect after creating a host management configuration, or if you receive an error, including an error about SSM Agent, see one of the following solutions:

• Solution B: No error, but still can't connect

• Solution C: Error about missing SSM Agent

Solution B: No error, but still can't connect

If you created the host management configuration, waited several minutes before trying to connect, and still can't connect, then you might need to manually apply the host management configuration to your instance. Use the following procedure to update a Quick Setup host management configuration and apply changes to an instance.

To update a host management configuration using Quick Setup

1. Open the AWS Systems Manager console at https://console.aws.amazon.com/systems-manager/.

2. In the navigation pane, choose Quick Setup.

3. In the Configurations list, choose the Host Management configuration you created.

4. Choose Actions, and then choose Edit configuration.

5. Near the bottom of the Targets section, under Choose how you want to target instances, choose Manual.

6. In the Instances section, choose the instance you created.

7. Choose Update.

Wait a few minutes for EC2 to refresh the Session Manager tab. If you still can't connect or if you receive an error, review the remaining solutions for this issue.

Solution C: Error about missing SSM Agent

If you weren't able to create a host management configuration by using Quick Setup, or if you received an error about SSM Agent not being installed, you may need to manually install SSM Agent on your instance. SSM Agent is Amazon software that enables Systems Manager to connect to your instance by using Session Manager. SSM Agent is installed by default on most Amazon Machine Images (AMIs). If your instance was created from a non-standard AMI or an older AMI, you might have to manually install the agent. For the procedure to install SSM Agent, see the following topic that corresponds to your instance operating system.

• Windows Server

• macOS

• AlmaLinux

• Amazon Linux 2 and AL2023

• Debian Server

• Oracle Linux

• Red Hat Enterprise Linux

• Rocky Linux

• Ubuntu Server

For issues with SSM Agent, see Troubleshooting SSM Agent.

No permission to start a session

Problem: You try to start a session, but the system tells you that you don't have the necessary permissions.

• Solution: A system administrator hasn't granted you AWS Identity and Access Management (IAM) policy permissions for starting Session Manager sessions. For information, see Control user session access to instances.

SSM Agent not online

Problem: You see a message on the Amazon EC2 instance Session Manager tab that states: "SSM Agent is not online. The SSM Agent was unable to connect to a Systems Manager endpoint to register itself with the service."

Solution: SSM Agent is Amazon software that runs on Amazon EC2 instances so that Session Manager can connect to them. If you see this error, SSM Agent is unable to establish a connection with the Systems Manager endpoint. Possible sources of the problem could be firewall restrictions, routing problems, or lack of internet connectivity. To resolve this issue, investigate network connectivity problems. For more information, see Troubleshooting SSM Agent and Troubleshooting managed node availability. For information about Systems Manager endpoints, see AWS Systems Manager endpoints and quotas in the AWS General Reference.

No permission to change session preferences

Problem: You try to update global session preferences for your organization, but the system tells you that you don't have the necessary permissions.

• Solution: A system administrator hasn't granted you IAM policy permissions for setting Session Manager preferences. For information, see Grant or deny a user permissions to update Session Manager preferences.

Managed node not available or not configured for Session Manager

Problem 1: You want to start a session on the Start a session console page, but a managed node isn't in the list.

• Solution A: The managed node you want to connect to might not have been configured for AWS Systems Manager. For more information, see Setting up Systems Manager unified console for an organization.

  Note

  If AWS Systems Manager SSM Agent is already running on a managed node when you attach the IAM instance profile, you might need to restart the agent before the instance is listed on the Start a session console page.

• Solution B: The proxy configuration you applied to the SSM Agent on your managed node might be incorrect. If the proxy configuration is incorrect, the managed node won't be able to reach the needed service endpoints, or the node might report as a different operating system to Systems Manager. For more information, see Configuring SSM Agent to use a proxy on Linux nodes and Configure SSM Agent to use a proxy for Windows Server instances.

Problem 2: A managed node you want to connect is in the list on the Start a session console page, but the page reports that "The instance you selected isn't configured to use Session Manager."

• Solution A: The managed node has been configured for use with the Systems Manager service, but the IAM instance profile attached to the node might not include permissions for the Session Manager tool. For information, see Verify or Create an IAM Instance Profile with Session Manager Permissions.

• Solution B: The managed node isn't running a version of SSM Agent that supports Session Manager. Update SSM Agent on the node to version 2.3.68.0 or later.

  Update SSM Agent manually on a managed node by following the steps in Manually installing and uninstalling SSM Agent on EC2 instances for Windows Server, Manually installing and uninstalling SSM Agent on EC2 instances for Linux, or Manually installing and uninstalling SSM Agent on EC2 instances for macOS, depending on the operating system.

  Alternatively, use the Run Command document AWS-UpdateSSMAgent to update the agent version on one or more managed nodes at a time. For information, see Updating the SSM Agent using Run Command.

  Tip

  To always keep your agent up to date, we recommend updating SSM Agent to the latest version on an automated schedule that you define using either of the following methods:

  • Run AWS-UpdateSSMAgent as part of a State Manager association. For information, see Walkthrough: Automatically update SSM Agent with the AWS CLI.

  • Run AWS-UpdateSSMAgent as part of a maintenance window. For information about working with maintenance windows, see Create and manage maintenance windows using the console and Tutorial: Create and configure a maintenance window using the AWS CLI.

• Solution C: The managed node can't reach the requisite service endpoints. You can improve the security posture of your managed nodes by using interface endpoints powered by AWS PrivateLink to connect to Systems Manager endpoints. The alternative to using interface endpoints is to allow outbound internet access on your managed nodes. For more information, see Use PrivateLink to set up a VPC endpoint for Session Manager.

• Solution D: The managed node has limited available CPU or memory resources. Although your managed node might otherwise be functional, if the node doesn't have enough available resources, you can't establish a session. For more information, see Troubleshooting an Unreachable Instance.

Session Manager plugin not found

To use the AWS CLI to run session commands, the Session Manager plugin must also be installed on your local machine. For information, see Install the Session Manager plugin for the AWS CLI.

Session Manager plugin not automatically added to command line path (Windows)

When you install the Session Manager plugin on Windows, the session-manager-plugin executable should be automatically added to your operating system's PATH environment variable. If the command failed after you ran it to check whether the Session Manager plugin installed correctly (aws ssm start-session --target instance-id), you might need to set it manually using the following procedure.

To modify your PATH variable (Windows)

1. Press the Windows key and enter environment variables.

2. Choose Edit environment variables for your account.

3. Choose PATH and then choose Edit.

4. Add paths to the Variable value field, separated by semicolons, as shown in this example: C:\existing\path;C:\new\path

   C:\existing\path represents the value already in the field. C:\new\path represents the path you want to add, as shown in the following example.

   • 64-bit machines: C:\Program Files\Amazon\SessionManagerPlugin\bin\

5. Choose OK twice to apply the new settings.

6. Close any running command prompts and re-open.

Session Manager plugin becomes unresponsive

During a port forwarding session, traffic might stop forwarding if you have antivirus software installed on your local machine. In some cases, antivirus software interferes with the Session Manager plugin causing process deadlocks. To resolve this issue, allow or exclude the Session Manager plugin from the antivirus software. For information about the default installation path for the Session Manager plugin, see Install the Session Manager plugin for the AWS CLI.

TargetNotConnected

Problem: You try to start a session, but the system returns the error message, "An error occurred (TargetNotConnected) when calling the StartSession operation: InstanceID isn't connected."

• Solution A: This error is returned when the specified target managed node for the session isn't fully configured for use with Session Manager. For information, see Setting up Session Manager.

• Solution B: This error is also returned if you attempt to start a session on a managed node that is located in a different AWS account or AWS Region.

• Solution C: SSM Agent on the managed node can't reach the Systems Manager service endpoints, so the node stops reporting as online. This is the most common cause when a node that worked before suddenly fails, or when a node in a private subnet never comes online. In this case, the SSM Agent log on the node typically contains an error similar to the following:

  SSM Agent unable to acquire credentials: unexpected error getting instance profile role credentials or calling UpdateInstanceInformation. Skipping default host management fallback: retrieved credentials failed to report to ssm. Error: RequestError: send request failed

  Check the following, starting from the VPC and working outward:

  • VPC endpoints: If the node reaches Systems Manager through AWS PrivateLink instead of a NAT gateway or an internet gateway, check the VPC endpoints. Verify that the required interface endpoints exist in the VPC. Confirm that their security groups allow inbound HTTPS (port 443) from the node's private IP. SSM Agent needs com.amazonaws.region.ssm and com.amazonaws.region.ssmmessages. SSM Agent versions earlier than 3.3.40.0 also need com.amazonaws.region.ec2messages. Version 3.3.40.0 and later use ssmmessages instead whenever it's available, and some newer AWS Regions don't offer an ec2messages endpoint. To list the endpoints in the node's VPC, run the following AWS CLI command:

    aws ec2 describe-vpc-endpoints \
        --filters Name=vpc-id,Values=vpc-id \
        --query "VpcEndpoints[].ServiceName"

    Replace vpc-id with the ID of the node's VPC. The output lists the service names of every endpoint in that VPC. Confirm that an endpoint ending in .ssm and one ending in .ssmmessages are present (plus one ending in .ec2messages for agent versions earlier than 3.3.40.0). Create any that are missing. For instructions on setting up the required interface endpoints, see Step 6: (Optional) Use AWS PrivateLink to set up a VPC endpoint for Session Manager.

  • VPC Block Public Access (BPA): If the node reaches Systems Manager through an internet gateway, check whether VPC BPA blocks its traffic. When VPC BPA is enabled in either block-ingress or block-bidirectional mode, a node that reaches the public Systems Manager endpoints through an internet gateway can no longer connect, and goes offline. To check whether VPC BPA is on in the AWS Region, run the following AWS CLI command:

    aws ec2 describe-vpc-block-public-access-options \
        --region region

    Replace region with the identifier of the node's AWS Region. In the output, check the InternetGatewayBlockMode field:

    • block-ingress: Route the node's traffic through a NAT gateway, or use interface VPC endpoints for Systems Manager (as described in the previous bullet).

    • block-bidirectional: VPC BPA also blocks traffic through NAT gateways in this mode. Use interface VPC endpoints for Systems Manager, or create a VPC BPA exclusion for the node's subnet or VPC. To allow outbound traffic only, create the exclusion in allow-egress mode:

      aws ec2 create-vpc-block-public-access-exclusion \
          --subnet-id subnet-id \
          --internet-gateway-exclusion-mode allow-egress \
          --region region

      Replace subnet-id with the ID of the node's subnet, and region with the identifier of the node's AWS Region. To exclude the whole VPC instead, use --vpc-id.

    For information about the VPC BPA modes and exclusions, see Block public access to VPCs and subnets in the Amazon VPC User Guide.

  • Security groups and network ACLs: Configure the security group attached to the node to allow outbound HTTPS (port 443) to the Systems Manager endpoints. Use the public ranges, or the interface endpoint ENIs if you are using AWS PrivateLink. Configure the network ACL for the subnet to allow both outbound HTTPS and the matching inbound ephemeral port range for return traffic.

  • VPC flow logs: If the first three checks look correct, enable VPC flow logs on the node's elastic network interface and start a connection. A REJECT entry on port 443 pinpoints which layer is actually dropping the traffic.

• Solution D: The managed node is running but under resource pressure, so SSM Agent can't keep its connection to Systems Manager open. CPU exhaustion, memory exhaustion, or a full root volume can all cause the node to stop reporting as online. Use CloudWatch or the operating system's own tools to confirm CPU, memory, and disk headroom, and free resources if any are at their limit.

• Solution E: SSM Agent isn't running on the node. On Linux & macOS, run systemctl status amazon-ssm-agent (or status amazon-ssm-agent on older init systems). On Windows Server, run Get-Service AmazonSSMAgent. If the agent is stopped, start it with the matching command for the operating system. For instructions and the complete list of operating-system-specific commands, see Checking SSM Agent status and starting the agent. For log file locations and further diagnostics, see Troubleshooting SSM Agent.

• Solution F: The node doesn't have an IAM instance profile attached, or the attached profile lacks the permissions SSM Agent needs to call Systems Manager. Attach a profile that includes the AmazonSSMManagedInstanceCore managed policy (or an equivalent custom policy) so the agent can register the node and keep the connection open. For instructions on attaching an instance profile, see Managing EC2 instances with Systems Manager.

Blank screen displays after starting a session

Problem: You start a session and Session Manager displays a blank screen.

• Solution A: This issue can occur when the root volume on the managed node is full. Due to lack of disk space, SSM Agent on the node stops working. To resolve this issue, use Amazon CloudWatch to collect metrics and logs from the operating systems. For information, see Collect metrics, logs, and traces with the CloudWatch agent in the Amazon CloudWatch User Guide.

• Solution B: A blank screen might display if you accessed the console using a link that includes a mismatched endpoint and Region pair. For example, in the following console URL, us-west-2 is the specified endpoint, but us-west-1 is the specified AWS Region.

  https://us-west-2.console.aws.amazon.com/systems-manager/session-manager/sessions?region=us-west-1

• Solution C: The managed node is connecting to Systems Manager using VPC endpoints, and your Session Manager preferences write session output to an Amazon S3 bucket or Amazon CloudWatch Logs log group, but an s3 gateway endpoint or logs interface endpoint doesn't exist in the VPC. An s3 endpoint in the format com.amazonaws.region.s3 is required if your managed nodes are connecting to Systems Manager using VPC endpoints, and your Session Manager preferences write session output to an Amazon S3 bucket. Alternatively, a logs endpoint in the format com.amazonaws.region.logs is required if your managed nodes are connecting to Systems Manager using VPC endpoints, and your Session Manager preferences write session output to a CloudWatch Logs log group. For more information, see Creating VPC endpoints for Systems Manager.

• Solution D: The log group or Amazon S3 bucket you specified in your session preferences has been deleted. To resolve this issue, update your session preferences with a valid log group or S3 bucket.

• Solution E: The log group or Amazon S3 bucket you specified in your session preferences isn't encrypted, but you have set the cloudWatchEncryptionEnabled or s3EncryptionEnabled input to true. To resolve this issue, update your session preferences with a log group or Amazon S3 bucket that is encrypted, or set the cloudWatchEncryptionEnabled or s3EncryptionEnabled input to false. This scenario is only applicable to customers who create session preferences using command line tools.

Managed node becomes unresponsive during long running sessions

Problem: Your managed node becomes unresponsive or crashes during a long running session.

Solution: Decrease the SSM Agent log retention duration for Session Manager.

To decrease the SSM Agent log retention duration for sessions

1. Locate the amazon-ssm-agent.json.template in the /etc/amazon/ssm/ directory for Linux, or C:\Program Files\Amazon\SSM for Windows.

2. Copy the contents of the amazon-ssm-agent.json.template to a new file in the same directory named amazon-ssm-agent.json.

3. Decrease the default value of the SessionLogsRetentionDurationHours value in the SSM property, and save the file.

4. Restart the SSM Agent.

An error occurred (InvalidDocument) when calling the StartSession operation

Problem: You receive the following error when starting a session by using the AWS CLI.

An error occurred (InvalidDocument) when calling the StartSession operation: Document type: 'Command' is not supported. Only type: 'Session' is supported for Session Manager.

Solution: The SSM document you specified for the --document-name parameter isn't a Session document. Use the following procedure to view a list of Session documents in the AWS Management Console.

To view a list of Session documents

1. Open the AWS Systems Manager console at https://console.aws.amazon.com/systems-manager/.

2. In the navigation pane, choose Documents.

3. In the Categories list, choose Session documents.