About the AWS-RunPatchBaselineWithHooks SSM document - Amazon Systems Manager
Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

About the AWS-RunPatchBaselineWithHooks SSM document

Amazon Systems Manager supports AWS-RunPatchBaselineWithHooks, a Systems Manager document (SSM document) for Patch Manager, a capability of Amazon Systems Manager. This SSM document performs patching operations on managed nodes for both security related and other types of updates.

AWS-RunPatchBaselineWithHooks differs from AWS-RunPatchBaseline in the following ways:

  • A wrapper documentAWS-RunPatchBaselineWithHooks is a wrapper for AWS-RunPatchBaseline and relies on AWS-RunPatchBaseline for some of its operations.

  • The Install operationAWS-RunPatchBaselineWithHooks supports lifecycle hooks that run at designated points during managed node patching. Because patch installations sometimes require managed nodes to reboot, the patching operation is divided into two events, for a total of three hooks that support custom functionality. The first hook is before the Install with NoReboot operation. The second hook is after the Install with NoReboot operation. The third hook is available after the reboot of the managed node.

  • No custom patch list supportAWS-RunPatchBaselineWithHooks doesn't support the InstallOverrideList parameter.

  • SSM Agent supportAWS-RunPatchBaselineWithHooks requires that SSM Agent 3.0.502 or later be installed on the managed node to patch.

When the document is run, it uses the patch baseline currently specified as the "default" for an operating system type if no patch group is specified. Otherwise, it uses the patch baselines that is associated with the patch group. For information about patch groups, see About patch groups.

You can use the document AWS-RunPatchBaselineWithHooks to apply patches for both operating systems and applications. (On Windows, application support is limited to updates for applications released by Microsoft.)

This document supports Linux, macOS, and Windows Server managed nodes. The document will perform the appropriate actions for each platform.

Linux

On Linux managed nodes, the AWS-RunPatchBaselineWithHooks document invokes a Python module, which in turn downloads a snapshot of the patch baseline that applies to the managed node. This patch baseline snapshot uses the defined rules and lists of approved and blocked patches to drive the appropriate package manager for each node type:

  • Amazon Linux 1, Amazon Linux 2, CentOS, Oracle Linux, and RHEL 7 managed nodes use YUM. For YUM operations, Patch Manager requires Python 2.6 or a later supported version (2.6 - 3.10).

  • RHEL 8 managed nodes use DNF. For DNF operations, Patch Manager requires a supported version of Python 2 or Python 3 (2.6 - 3.10). (Neither version is installed by default on RHEL 8. You must install one or the other manually.)

  • Debian Server, Raspberry Pi OS, and Ubuntu Server instances use APT. For APT operations, Patch Manager requires a supported version of Python 3 (3.0 - 3.10).

  • SUSE Linux Enterprise Server managed nodes use Zypper. For Zypper operations, Patch Manager requires Python 2.6 or a later supported version (2.6 - 3.10).

macOS

On macOS managed nodes, the AWS-RunPatchBaselineWithHooks document invokes a Python module, which in turn downloads a snapshot of the patch baseline that applies to the managed node. Next, a Python subprocess invokes the CLI on the node to retrieve the installation and update information for the specified package managers and to drive the appropriate package manager for each update package.

Windows Server

On Windows Server managed nodes, the AWS-RunPatchBaselineWithHooks document downloads and invokes a PowerShell module, which in turn downloads a snapshot of the patch baseline that applies to the managed node. This patch baseline snapshot contains a list of approved patches that is compiled by querying the patch baseline against a Windows Server Update Services (WSUS) server. This list is passed to the Windows Update API, which controls downloading and installing the approved patches as appropriate.

Each snapshot is specific to an Amazon Web Services account, patch group, operating system, and snapshot ID. The snapshot is delivered through a presigned Amazon Simple Storage Service (Amazon S3) URL, which expires 24 hours after the snapshot is created. After the URL expires, however, if you want to apply the same snapshot content to other managed nodes, you can generate a new presigned Amazon S3 URL up to three days after the snapshot was created. To do this, use the get-deployable-patch-snapshot-for-instance command.

After all approved and applicable updates have been installed, with reboots performed as necessary, patch compliance information is generated on an managed node and reported back to Patch Manager.

Note

If the RebootOption parameter is set to NoReboot in the AWS-RunPatchBaselineWithHooks document, the managed node isn't rebooted after Patch Manager runs. For more information, see Parameter name: RebootOption.

For information about viewing patch compliance data, see About patch compliance.

AWS-RunPatchBaselineWithHooks operational steps

When the AWS-RunPatchBaselineWithHooks runs, the following steps are performed:

  1. Scan - A Scan operation using AWS-RunPatchBaseline is run on the managed node, and a compliance report is generated and uploaded.

  2. Verify local patch states - A script is run to determine what steps will be performed based on the selected operation and Scan result from Step 1.

    1. If the selected operation is Scan, the operation is marked complete. The operation concludes.

    2. If the selected operation is Install, Patch Manager evaluates the Scan result from Step 1 to determine what to run next:

      1. If no missing patches are detected, and no pending reboots required, the operation proceeds directly to the final step (Step 8), which includes a hook you have provided. Any steps in between are skipped.

      2. If no missing patches are detected, but there are pending reboots required and the selected reboot option is NoReboot, the operation proceeds directly to the final step (Step 8), which includes a hook you have provided. Any steps in between are skipped.

      3. Otherwise, the operation proceeds to the next step.

  3. Pre-patch hook operation - The SSM document you have provided for the first lifecycle hook, PreInstallHookDocName, is run on the managed node.

  4. Install with NoReboot - An Install operation with the reboot option of NoReboot using AWS-RunPatchBaseline is run on the managed node, and a compliance report is generated and uploaded.

  5. Post-install hook operation - The SSM document you have provided for the second lifecycle hook, PostInstallHookDocName, is run on the managed node.

  6. Verify reboot - A script runs to determine whether a reboot is needed for the managed node and what steps to run:

    1. If the selected reboot option is NoReboot, the operation proceeds directly to the final step (Step 8), which includes a hook you have provided. Any steps in between are skipped.

    2. If the selected reboot option is RebootIfNeeded, Patch Manager checks whether there are any pending reboots required from the inventory collected in Step 4. This means that the operation continues to Step 7 and the managed node is rebooted in either of the following cases:

      1. Patch Manager installed one or more patches. (Patch Manager doesn't evaluate whether a reboot is required by the patch. The system is rebooted even if the patch doesn't require a reboot.)

      2. Patch Manager detects one or more patches with a status of INSTALLED_PENDING_REBOOT during the Install operation. The INSTALLED_PENDING_REBOOT status can mean that the option NoReboot was selected the last time the Install operation was run."

      If no patches meeting these criteria are found, the managed node patching operation is complete, and the operation proceeds directly to the final step (Step 8), which includes a hook you have provided. Any steps in between are skipped.

  7. Reboot and report - An installation operation with the reboot option of RebootIfNeeded runs on the managed node using AWS-RunPatchBaseline, and a compliance report is generated and uploaded.

  8. Post-reboot hook operation - The SSM document you have provided for the third lifecycle hook, OnExitHookDocName, is run on the managed node.

For a Scan operation, if Step 1 fails, the process of running the document stops and the step is reported as failed, although subsequent steps are reported as successful.

For an Install operation, if any of the aws:runDocument steps fail during the operation, those steps are reported as failed, and the operation proceeds directly to the final step (Step 8), which includes a hook you have provided. Any steps in between are skipped. This step is reported as failed, the last step reports the status of its operation result, and all steps in between are reported as successful.

AWS-RunPatchBaselineWithHooks parameters

AWS-RunPatchBaselineWithHooks supports six parameters.

The Operation parameter is required.

The RebootOption, PreInstallHookDocName, PostInstallHookDocName, and OnExitHookDocName parameters are optional.

Snapshot-ID is technically optional, but we recommend that you supply a custom value for it when you run AWS-RunPatchBaselineWithHooks outside of a maintenance window. Let Patch Manager supply the value automatically when the document is run as part of a maintenance window operation.

Parameter name: Operation

Usage: Required.

Options: Scan | Install.

Scan

When you choose the Scan option, the systems uses the AWS-RunPatchBaseline document to determine the patch compliance state of the managed node and reports this information back to Patch Manager. Scan doesn't prompt updates to be installed or managed nodes to be rebooted. Instead, the operation identifies where updates are missing that are approved and applicable to the node.

Install

When you choose the Install option, AWS-RunPatchBaselineWithHooks attempts to install the approved and applicable updates that are missing from the managed node. Patch compliance information generated as part of an Install operation doesn't list any missing updates, but might report updates that are in a failed state if the installation of the update didn't succeed for any reason. Whenever an update is installed on a managed node, the node is rebooted to ensure the update is both installed and active. (Exception: If the RebootOption parameter is set to NoReboot in the AWS-RunPatchBaselineWithHooks document, the managed node isn't rebooted after Patch Manager runs. For more information, see Parameter name: RebootOption.)

Note

If a patch specified by the baseline rules is installed before Patch Manager updates the managed node, the system might not reboot as expected. This can happen when a patch is installed manually by a user or installed automatically by another program, such as the unattended-upgrades package on Ubuntu Server.

Parameter name: Snapshot ID

Usage: Optional.

Snapshot ID is a unique ID (GUID) used by Patch Manager to ensure that a set of managed nodes that are patched in a single operation all have the exact same set of approved patches. Although the parameter is defined as optional, our best practice recommendation depends on whether or not you're running AWS-RunPatchBaselineWithHooks in a maintenance window, as described in the following table.

AWS-RunPatchBaselineWithHooks best practices
Mode Best practice Details
Running AWS-RunPatchBaselineWithHooks inside a maintenance window Don't supply a Snapshot ID. Patch Manager will supply it for you.

If you use a maintenance window to run AWS-RunPatchBaselineWithHooks, you shouldn't provide your own generated Snapshot ID. In this scenario, Systems Manager provides a GUID value based on the maintenance window execution ID. This ensures that a correct ID is used for all the invocations of AWS-RunPatchBaselineWithHooks in that maintenance window.

If you do specify a value in this scenario, note that the snapshot of the patch baseline might not remain in place for more than 3 days. After that, a new snapshot will be generated even if you specify the same ID after the snapshot expires.

Running AWS-RunPatchBaselineWithHooks outside of a maintenance window Generate and specify a custom GUID value for the Snapshot ID.¹

When you aren't using a maintenance window to run AWS-RunPatchBaselineWithHooks, we recommend that you generate and specify a unique Snapshot ID for each patch baseline, particularly if you're running the AWS-RunPatchBaselineWithHooks document on multiple managed nodes in the same operation. If you don't specify an ID in this scenario, Systems Manager generates a different Snapshot ID for each managed node the command is sent to. This might result in varying sets of patches being specified among the nodes.

For instance, say that you're running the AWS-RunPatchBaselineWithHooks document directly through Run Command, a capability ofAmazon Systems Manager, and targeting a group of 50 managed nodes. Specifying a custom Snapshot ID results in the generation of a single baseline snapshot that is used to evaluate and patch all the managed nodes, ensuring that they end up in a consistent state.

¹ You can use any tool capable of generating a GUID to generate a value for the Snapshot ID parameter. For example, in PowerShell, you can use the New-Guid cmdlet to generate a GUID in the format of 12345699-9405-4f69-bc5e-9315aEXAMPLE.

Parameter name: RebootOption

Usage: Optional.

Options: RebootIfNeeded | NoReboot

Default: RebootIfNeeded

Warning

The default option is RebootIfNeeded. Be sure to select the correct option for your use case. For example, if your managed nodes must reboot immediately to complete a configuration process, choose RebootIfNeeded. Or, if you need to maintain managed node availability until a scheduled reboot time, choose NoReboot.

Important

We don’t recommend using Patch Manager for patching cluster instances in Amazon EMR (previously called Amazon Elastic MapReduce). In particular, don’t select the RebootIfNeeded option for the RebootOption parameter. (This option is available in the SSM Command documents for patching AWS-RunPatchBaseline, AWS-RunPatchBaselineAssociation, and AWS-RunPatchBaselineWithHooks.)

The underlying commands for patching using Patch Manager use yum and dnf commands. Therefore, the operations result in incompatibilities because of how packages are installed. For information about the preferred methods for updating software on Amazon EMR clusters, see Using the default AMI for Amazon EMR in the Amazon EMR Management Guide.

RebootIfNeeded

When you choose the RebootIfNeeded option, the managed node is rebooted in either of the following cases:

  • Patch Manager installed one or more patches.

    Patch Manager doesn't evaluate whether a reboot is required by the patch. The system is rebooted even if the patch doesn't require a reboot.

  • Patch Manager detects one or more patches with a status of INSTALLED_PENDING_REBOOT during the Install operation.

    The INSTALLED_PENDING_REBOOT status can mean that the option NoReboot was selected the last time the Install operation was run.

    Note

    Patches installed outside of Patch Manager are never given a status of INSTALLED_PENDING_REBOOT.

Rebooting managed nodes in these two cases ensures that updated packages are flushed from memory and keeps patching and rebooting behavior consistent across all operating systems.

NoReboot

When you choose the NoReboot option, Patch Manager doesn't reboot a managed node even if it installed patches during the Install operation. This option is useful if you know that your managed nodes don't require rebooting after patches are applied, or you have applications or processes running on a node that shouldn't be disrupted by a patching operation reboot. It's also useful when you want more control over the timing of managed node reboots, such as by using a maintenance window.

Note

If you choose the NoReboot option and a patch is installed, the patch is assigned a status of InstalledPendingReboot. The managed node itself, however, is marked as Non-Compliant. After a reboot occurs and a Scan operation is run, the node status is updated to Compliant.

Patch installation tracking file: To track patch installation, especially patches that were installed since the last system reboot, Systems Manager maintains a file on the managed node.

Important

Don't delete or modify the tracking file. If this file is deleted or corrupted, the patch compliance report for the managed node is inaccurate. If this happens, reboot the node and run a patch Scan operation to restore the file.

This tracking file is stored in the following locations on your managed nodes:

  • Linux operating systems:

    • /var/log/amazon/ssm/patch-configuration/patch-states-configuration.json

    • /var/log/amazon/ssm/patch-configuration/patch-inventory-from-last-operation.json

  • Windows Server operating system:

    • C:\ProgramData\Amazon\PatchBaselineOperations\State\PatchStatesConfiguration.json

    • C:\ProgramData\Amazon\PatchBaselineOperations\State\PatchInventoryFromLastOperation.json

Parameter name: PreInstallHookDocName

Usage: Optional.

Default: AWS-Noop.

The value to provide for the PreInstallHookDocName parameter is the name or Amazon Resource Name (ARN) of an SSM document of your choice. You can provide the name of an Amazon managed document or the name or ARN of a custom SSM document that you have created or that has been shared with you. (For an SSM document that has been shared with you from a different Amazon Web Services account, you must specify the full resource ARN, such as arn:aws:ssm:us-east-2:123456789012:document/MySharedDocument.)

The SSM document you specify is run before the Install operation and performs any actions supported by SSM Agent, such as a shell script to check application health check before patching is performed on the managed node. (For a list of actions, see Command document plugin reference). The default SSM document name is AWS-Noop, which doesn't perform any operation on the managed node.

For information about creating a custom SSM document, see Creating SSM document content.

Parameter name: PostInstallHookDocName

Usage: Optional.

Default: AWS-Noop.

The value to provide for the PostInstallHookDocName parameter is the name or Amazon Resource Name (ARN) of an SSM document of your choice. You can provide the name of an Amazon managed document or the name or ARN of a custom SSM document that you have created or that has been shared with you. (For an SSM document that has been shared with you from a different Amazon Web Services account, you must specify the full resource ARN, such as arn:aws:ssm:us-east-2:123456789012:document/MySharedDocument.)

The SSM document you specify is run after the Install with NoReboot operation and performs any actions supported by SSM Agent, such as a shell script for installing third party updates before reboot. (For a list of actions, see Command document plugin reference). The default SSM document name is AWS-Noop, which doesn't perform any operation on the managed node.

For information about creating a custom SSM document, see Creating SSM document content.

Parameter name: OnExitHookDocName

Usage: Optional.

Default: AWS-Noop.

The value to provide for the OnExitHookDocName parameter is the name or Amazon Resource Name (ARN) of an SSM document of your choice. You can provide the name of an Amazon managed document or the name or ARN of a custom SSM document that you have created or that has been shared with you. (For an SSM document that has been shared with you from a different Amazon Web Services account, you must specify the full resource ARN, such as arn:aws:ssm:us-east-2:123456789012:document/MySharedDocument.)

The SSM document you specify is run after the managed node reboot operation and performs any actions supported by SSM Agent, such as a shell script to verify node health after the patching operation is complete. (For a list of actions, see Command document plugin reference). The default SSM document name is AWS-Noop, which doesn't perform any operation on the managed node.

For information about creating a custom SSM document, see Creating SSM document content.