HP Clustered File System 3.6.
Version: HP Clustered File System 3.6.1 Description This document describes the HP Clustered File System 3.6.1 release. Supersedes HP Clustered File System 3.6.0 Operating systems NOTE: This section applies to the software-only version of HP Clustered File System 3.6.1. The supported operating systems for HP Clustered File System 3.6.
• Support for HP MSA 2xxx arrays. • Installation of the Microsoft SNMP service is no longer required to enable HP Clustered File System SNMP traps. It is still required to enable the HP Clustered File System extension agent. New features in FS Option for Windows This release includes the following new FS Option for Windows features: • Role-based security.
• Defect 19672. After the Configure Cluster GUI was opened, the “Starting Configuration Utility” status popup also remained open. • Defect 19659. A dynamic volume was inaccessible after a cluster transition occurred while the volume was being deported. • Defect 19704. Some non-cached file extending writes could result in an inconsistent internal state and cause a panic. • Defect 19716. An invalid attempt to release a resource caused a kernel fault. • Defect 19756. Alerts in HP Clustered File System 3.4 and 3.
New features in 3.6.0 HP Clustered File System 3.6.0 includes the following new features. Security features: • Single sign-on. HP Clustered File System now uses AD single sign-on semantics to authorize users attempting to access the cluster. • Role-Based Security Control. You can create roles, define the operations that are allowed or denied for each role, and add specific user or group accounts to the roles. • Auditing of all login attempts and user-initiated actions that change the cluster configuration.
• Networks can now be excluded from HP Clustered File System administrative traffic. The mx netif command has new options for specifying whether a network should be used for administrative traffic and the option and output from the mx netif status command have changed. • HP Clustered File System notifiers have been replaced by the new script event notifier.
• Alternate Data Streams (ADS). • Byte-range locking. PSFS supports both local-node and cross-node byte-range locking. • Cached and non-cached file access. • Case-insensitive naming. • Case-sensitive naming. • Content indexing. PSFS support is the same as the FAT filesystem. (See “Unsupported Features” for more information.) • Directory Change Notification (DCN). PSFS supports both local-node and cross-node Change Notification. • Disk quotas. • 8.3 short file names and name tunneling.
• File encryption and compression. This behavior is consistent with the FAT filesystem. • Hard links. • System ACLs and filesystem auditing. • NTFS change journal. As a consequence, file replication services (FRS) cannot be used. Content indexing is also affected. The content indexing service uses the NTFS change journal to determine when to update the indices. In the PSFS filesystem, updates to the indices are done less frequently. • Object IDs. • Online defragmentation. • Memory-mapped sparse files.
• • • • • the local Explorer window. You will need to close and reopen Explorer to see the renamed directory. Memory-mapped files are not coherent across nodes. Because PSFS is a shared filesystem, the same file can be mapped into memory on more than one node. When a change is made to a file on a remote node, the local node does not see the change immediately. This occurs because the operating system on the remote node may not immediately write the change to disk.
from this service. If you do not need the Remote Storage Service, it can be disabled in the Services Control Panel applet. Windows applications You may notice the following differences from NTFS filesystems when running applications on PSFS filesystems: • Certain Windows applications will disable NTFS-related features when run on a PSFS filesystem. For example, older versions of Robocopy will not copy security descriptors. (As a workaround, use Xcopy to copy the security descriptors.
Issue Description 5519 SNMP dies on QLogic SANbox2 FC switches In a cluster with a QLogic SANbox2 FibreChannel switch, SNMP can die periodically. When this occurs, servers cannot be fenced, and servers that have already been fenced cannot unfence themselves. To recover, go to Admin mode on the switch and then run the hotreset command: admin start admin> hotreset The hotreset command resets the switch. (Before the switch is reset, all running applications save their FibreChannel state.
Issue Description 6364 Server’s IP address may change if assigned with DHCP When DHCP or DHCP reservations are used to assign IP addresses to servers, the IP address can change when the server is rebooted. The DHCP server may be slow to respond to the reboot, resulting in the server temporarily using a random automatic IP address until the DHCP address is acquired.
Issue Description 8290 Virtual host can cause route change When a virtual host is instantiated it uses the network mask of the interface upon which it is activated. This can cause a unexpected route change if the virtual host address was intended for some other interface on another network or if the virtual host address is mistyped. For example, the following command adds the 10.10.1.100 address on the 10.11.1.1 interface. mx vhost add 10.10.1.100 10.11.1.1 If the 10.11.1.
Issue Description Cluster may not start when fencing method is changed If the fencing method is changed and all servers in the cluster were not shut down cleanly, the cluster may not be able to start successfully. To check for this condition and correct it if necessary, first run the mpdump command with no options. mpdump Membership Partition Host Registry (Version 3): Host ID: 10.10.180.5 fencestatus=0 fencetype=0 Fence ID:21:00:00:e0:8b:0c:12:1f::10.10.11.221 Host ID: 10.10.180.
Issue Description Shadow copy option is available for snapclones 13275 When you create a snapclone on the Create Snapshot window, the option to “Share as Shadow Copy of Shared Folder” is available; however, a shadow copy cannot be created from a snapclone and the option is unused.
Issue Description lanman IRPStackSize is too small After installing applications that use the IRP stack (such as some AntiVirus software), you may see error messages indicating that the value of the lanman IRPStackSize parameter needs to be increased. This problem has also been seen when upgrading HP Clustered File System from a release earlier than 3.4. (HP Clustered File System uses an additional IRP stack in release 3.4 and later.
Issue Description Removing the Microsoft SNMP service causes HP Clustered File System SNMP agent to fail If the Microsoft SNMP service is uninstalled, the HP Clustered File System SNMP extension agent will fail to send traps. This problem occurs because registry entries for the HP Clustered File System SNMP extension agent are removed when the Microsoft SNMP service is uninstalled.
Issue Description mxcheck error message regarding conflicting network addresses is unclear When mxcheck detects that multiple interfaces have the same network address, it reports the following: 19168 19178 Failure: Check IP network and interface assignments (networks) Interface 10006 shares a network address with interface 10008, Matrix Server requires each interface to have a unique IP network address.
Issue Description mx check command incorrectly reports that fencing is not configured This situation can occur when mx commands are used to configure the cluster, including the fencing method. The fencing configuration file, fence.conf, is included in the files exported to the other nodes; however, it might not exist on the node where the configuration was created. The mx check command reports 20987 the error because it cannot find the fence.conf file. The missing fence.
HP Clustered File System FS Option for Windows The following issues affect FS Option for Windows. Issue Description 4507 File shares with duplicate names are problematic If a Cluster File Share and a Virtual File Share are created with the same name, the automated monitors can become confused about which File Share they are monitoring. This situation can cause unexpected failover behavior or can cause a File Share to point to the wrong resource. Workaround.
Issue Description 9564 Creating Virtual File Server with ALL_SERVERS option can cause confusion When creating Virtual File Servers with the ALL_SERVERS command-line option, you should be aware of the following: • When a Virtual File Server is created with the command-line ALL_SERVERS option, any servers added to the cluster at a later time will automatically be added to the Virtual File Server.
Issue Description Cannot remove share that is a DFS root target If you have a CIFS share that is also a DFS root target, the DFS service must be stopped before the share can be removed. Occasionally it is necessary to remove shares managed by FS Option for Windows, such as at shutdown time. If FS Option for Windows is unable to stop the DFS service, it will not be able to remove such shares and they will remain after the product has stopped.
Environment issues The following environment issues affect the operation of HP Clustered File System. Issue Description 5571 Emulex HBAnyware fails to install HP Clustered File System requires that the HBAnyware package be installed on systems with Emulex HBAs. The HBAnyware installation sometimes fails on Windows 2003 systems. If you experience this problem, reboot the system and then retry the installation.
