Home » Security hardening for openstack-ansible
The openstack-ansible-security role is highly configurable and it ships with reasonable defaults that provide a balance of enhanced security with limited interruptions to production environments. However, some deployers may need to adjust the default configurations to fit the particular needs of their OpenStack environment.
All of the default configurations are found within defaults/main.yml in the role itself.
NOTE: Deployers are strongly urged to test this role and any configuration changes in a non-production, test environment.
The STIG sets requirements for integrity monitoring of the system and the role will install AIDE to meet these requirements. However, Ubuntu’s default configuration will cause AIDE to search through container filesystems and this will cause a serious performance hit to disk I/O for very long periods.
By default, this role excludes some directories that cause AIDE to spend a very long time indexing the disk. Additional directories can be added to the exclusion list, but the directories currenty listed should not be removed. Deployers can add directories to aide_exclude_dirs.
The first AIDE database initialization can consume lots of CPU time and I/O resources. By default, the role won’t run the database initialization after the role is applied to avoid causing performance degradation. The database will be initialized automatically when the AIDE cron job runs the next day. To force the initialization to run as soon as the role finishes running, change the the following variable to true:
initialize_aide: true
The audit daemon is required by the STIG and it provides useful logging of critical events on a Linux server.
The openstack-ansible-security role creates a file containing audit rules for hosts.
Each group of rules are controlled by Ansible variables that begin with security_audit_. To omit a set of rules on a host, set the variable to no. To include a set of rules on a host, set the variable to yes.
For example, setting security_audit_filesystem_mounts to yes will ensure that the rules for auditing filesystem mounts are included on each host. Setting security_audit_filesystem_mounts to no will omit that group of rules on each host.
To review the full list of rules and variables, refer to templates/osas-auditd.j2.
There are several configurations for auditd which are critical for deployers to review in detail. The options beneath the ## Auditd configuration comment will change how auditd handles log files and what it should do in case of emergencies.
NOTE: Some of these configuration options can cause serious issues on production systems, ranging from a reduction in security to servers going offline unexpectedly. There is extensive documentation in the developer notes for each STIG requirement.
The STIG sets requirements for various authentication-related security controls, including password complexity, password aging/locking, and simultaneous logins. All of these can cause issues on production systems and some cannot be met on Ubuntu systems due to the lack of certain packages.
Deployers are urged to review each item under the ## Authentication comment in defaults/main.yml as well as the developer notes for each requirement. The notes explain the potential pitfalls from each configuration item and they provide alternatives when a configuration isn’t directly available.
The fail2ban service is installed to meet some requirements around failed login attempts. The STIG requires pam_faillock, but that module isn’t available in Ubuntu 14.04.
To opt-in for the fail2ban service to be installed, set install_fail2ban to yes and set an appropriate time for bans with fail2ban_bantime. See the notes for V-38501 for more details.
Certain kernel modules are restricted by the STIG because they can become a security threat to a server. The Ansible tasks will disable most of these variables in accordance with the STIG. These changes are controlled by Ansible variables matching the pattern security_disable_module_MODULENAME. Refer to defaults/main.yml for a full list of these variables.
A setting of yes means that the module will be disabled on the next boot and a setting of no means that the state of the module will not be changed.
All of the defaults are set in accordance with the STIG’s requitements with the exception of the usb_storage kernel module. This module is used frequently with external hard drives, USB sticks, and with some IPMI implementations. Deployers who wish to follow the STIG guidelines will need to set usb_storage to yes so that the usb_storage module is disabled on the next boot.
The STIG requires that TCP SYN cookies enabled by default to protect against certain types of attacks, like SYN floods. This can cause issues in some environments with busy load balancers. Deployers should review the notes for V-38539 for more details.
Also, the STIG requires IPv6 support to be fully disabled, and this could cause issues for production systems. The role will not disable IPv6 by default, but deployers can adjust this by changing disable_ipv6 to yes.
Core dumps are also disabled by default in the openstack-ansible-security role.
The STIG requires that SELinux is in enforcing mode to provide additional security against attacks. The security role will enable AppArmor on Ubuntu systems to meet this requirement.
For more information on how these changes are applied, refer to the documentation for V-51337.
Deployers are strongly urged to configure an address to receive the root user’s email on various hosts. This is done with the root_forward_email variable.
The STIG requires that a valid user receives the email in case of errors or a security issue.
The STIG has recommendations for which services shouldn’t be running and which packages shouldn’t be installed. These removals can be configured to meet the requirements of the deployer.
By default, the role will disable any services that are recommended to be disabled by the STIG. These changes are controlled by Ansible variables that match the security_disable_SERVICENAME pattern. Review these variables in defaults/main.yml for more details.
A setting of yes for a service will cause the service to be disabled in accordance to the STIG’s requirements.
A setting of no causes the role to ignore the service entirely. If the service is running, it will remain running. If the service is stopped, it will remain stopped.
The STIG requires that some packages are completely removed from the server. By default, the role will remove the packages in accordance with the STIG’s requirements. These changes are controlled by Ansible variables that match the security_remove_SERVICENAME pattern. Review these variables in defaults/main.yml for more details.
A setting of yes for a service will cause the package that contains the service to be removed from the system. If the service happens to be running at the time, it will be stopped by apt.
A setting of no for a service will cause the role to ignore the package that contains the service. If the package is installed, it will be left installed.
The STIG has some requirements for ssh server configuration and these requirements are applied by default by the role. To opt-out or change these requirements, see the section under the ## SSH configuration comment in defaults/main.yml.
NOTE: There is one deviation from the STIG for the PermitRootLogin configuration option. The STIG requires that direct root logins are disabled, and this is the recommended setting for secure production environments. However, this can cause problems in some existing environments and the default for the role is to set it to yes (direct root logins allowed).
Reliable time synchronization is a requirement in the STIG and the chrony package will be installed to handle NTP for systems secured with the openstack-ansible-security role.
The default settings will work for most environments, but some deployers may prefer to use NTP servers which are geographically closer to their servers. Also, the default configuration allows RFC1918 addresses to reach the NTP server running on each host. That could be reduced to 127.0.0.1/32 for greater security.
Certain umask adjustments are required by the STIG, but these can cause problems with production systems. The requirements are commented out within defaults/main.yml and can be applied by uncommenting the variables that start with umask_*. There is extensive documentation available within the developer notes for each STIG requirement.