- The speed of fencing is critical because otherwise, broken nodes have more time to corrupt data. Prior to this update, the operation of the fence_vmware_soap fencing agent was slow and could corrupt data when used on the VMWare vSphere platform with hundreds of virtual machines. This update fixes a problem with virtual machines that do not have a valid UUID, which can be created during failed P2V (Physical-to-Virtual) processes. Now, the fencing process is also much faster and it does not terminate if a virtual machines without an UUID is encountered.
- The fence_rhevm fencing agent uses the Red Hat Enterprise Virtualization API to check the power status ("on" or "off") of a virtual machine. In addition to the "up" and "down" states, the API includes number of other states. Previously, only if the machine was in the "up" state, the "on" power status was returned. The "off" status was returned for all other states even if the machine was running. This allowed for successful fencing before the machine was really powered off. With this update, the fence_rhevm agent detects the power status of a cluster node more conservatively, and the "off" status is returned only if the machine is actually powered off, that is in the "down" state.
- Previously, the fence_soap_vmware fence agent was not able to work with more than one hundred machines in a cluster. Consequently, fencing a cluster node running in a virtual machine on VMWare with the fence_soap_vmware fence agent failed with the "KeyError: 'config.uuid'" error message. With this update, the underlying code has been fixed to support fencing on such clusters.
- Previously, the fence_ipmilan agent failed to handle passwd_script argument values that contained space characters. Consequently, it was impossible to use a password script that required additional parameters. This update ensures that fence_ipmilan accepts and properly parses values for the passwd_script argument with spaces.
- Previously, the fence_vmware_soap fence agent did not expose the proper virtual machine path for fencing. With this update, fence_vmware_soap has been fixed to support this virtual machine identification.
- Previously, certain fence agents did not generate correct metadata output. As a result, it was not possible to use the metadata for automatic generation of manual pages and user interfaces. With this update, all fence agents generate their metadata as expected.
- Possible buffer overflow and null dereference defects were found by automatic tools. With this update, these problems have been fixed.
- Fence agents that use an identity file for SSH terminated unexpectedly when a password was expected but was not provided. This bug has been fixed and proper error messages are returned in the described scenario.
- The fence_ipmilan fence agent did not respect the power_wait option and did not wait after sending the power-off signal to a device. Consequently, the device could terminate its shutdown sequence. This bug has been fixed and fence_ipmilan now waits before shutting down a machine as expected.
- The fence_scsi agent creates the fence_scsi.dev file that contains a list of devices that the node registered with during an unfence operation. This file was unlinked for every unfence action. Consequently, if multiple fence device entries were used in the cluster.conf file, fence_scsi.dev only contained the devices that the node registered with during the most recent unfence action. Now, instead of the unlink call, if the device currently being registered does not exists in fence_scsi.dev, it is added to the file.
- If the "delay" option was set to more than 5 seconds while a fence device was connected via the telnet_ssl utility, the connection timed out and the fence device failed. Now, the "delay" option is applied before the connection is opened, thus fixing this bug.
- Previously, XML metadata returned by a fence agent incorrectly listed all attributes as "unique". This update fixes this problem and the attributes are now marked as unique only when this information is valid.
- This update fixes a typographical error in an error message in the fence_ipmilan agent.
- Prior to this update, the fence agent for IPMI (Intelligent Platform Management Interface) could return an invalid return code when the "-M cycle" option was used. This invalid return code could cause invalid interpretation of a fence action, eventually causing the cluster to become unresponsive. This bug has been fixed and only predefined return codes are now returned in the described scenario.
- Previously, the fence_brocade fence agent did not distinguish the "action" option from the standard "option" option. Consequently, the "action" option was ignored and the node was always fenced. This bug has been fixed and both options are now properly recognized and acted upon.
- This updates adds the feature to access Fujitsu RSB fencing device using secure shell.