Xen Cloud Platform 1.6 Release Notes
See License information for software hosted on xen.org.
For minimum host requirements for running XCP, see XCP Release Features.
New Features and Improvements Since XCP 1.5
XCP 1.6 includes the following new features and ongoing improvements:
Storage XenMotion allows running VMs to be moved from one host to another. This includes the case where (a) VMs are not located on storage
shared between the hosts and (b) hosts are not in the same resource pool. This enables system administrators to:
- Rebalance or move VMs between XCP pools – for example promoting a VM from a development environment to a production environment;
- Perform software maintenance – for example upgrading or updating standalone XCP hosts without VM downtime;
- Perform hardware maintenance – for example upgrading standalone XCP host hardware without VM downtime;
- Reduce deployment costs by using local storage.
Live VDI Migration:
Live VDI Migration allows system administrators to relocate a VM's Virtual Disk Image (VDI) without shutting down the VM. This enables system administrators to:
- Move a VM from cheap, local storage to fast, resilient, array-backed storage;
- Move a VM from a development to a production environment;
- Move between tiers of storage when a VM is limited by storage capacity;
- Perform storage array upgrades.
- Link Aggregation Control Protocol (LACP) support: enables the use of industry-standard network bonding features to provide
fault-tolerance and load balancing of network traffic.
- Source Load Balancing (SLB) improvements: allows up to 4 NICs to be used in an active-active bond. This improves total
network throughput and increases fault tolerance in the event of hardware failures. The SLB balancing algorithm has been modified
to reduce load on switches in large deployments.
- Multi-Tenancy improvements: allows system administrators to restrict a VM to send and receive traffic on a specific MAC address
and a number of IPv4 or IPv6 addresses, without relying on VLANs and switch management software. When these extensions are deployed VMs
cannot impersonate any other VM, or intercept traffic intended for any other VM. This increases security in environments where VMs cannot
be fully trusted.
- VLAN Scalability improvements: removes a previous limitation which caused VM deployment delays when large numbers of VLANs were
in use. This improvement enables administrators using XCP 1.6 to deploy hundreds of VLANs in a XCP pool quickly.
- Emergency Network Reset: provides a simple mechanism to recover and reset a host's networking, allowing system administrators to
revert XCP hosts to a known good networking state.
- IPv6 Guest Support: enables the use of IPv6 addresses within guests allowing network administrators to plan for network growth.
Enhanced Guest OS Support: Newly Supported Guests
- Ubuntu 12.04
- CentOS 5.7, 1.5, 6.1, 6.2
- Red Hat Enterprise Linux 5.7, 6.1, 6.2
- Oracle Enterprise Linux 5.7, 6.1, 6.2
- Simplified mechanism to adjust Control Domain ("dom0") vCPUs.
- Updated Open vSwitch: v1.4.2 provides stability and performance improvements.
- Integrated StorageLink (iSL) support for EMC VNX series arrays.
- GPU Pass-through: support for up to 4 GPUs per host.
- Interoperability extensions for 3rd Party Tools: additional asynchronous XenAPI C language binding, Workload Balancing (WLB) extensions and general improvements.
- Support for hypervisor monitoring (vhostmd) allows SAP software to run inside a XCP VM.
Community changelog for ISP's and XCP Power Users & Release Notes Errata
Please check out the Community Section of the Release Notes,
which can be edited by everyone who has signed up on the wiki. You can cover anything we missed here in error, or highlight
errors in the release notes.
We will regularly review the Community Section of the Release Notes
and migrate content to this page.
XCP 1.6 Beta Releases History
|XCP 1.6 Final
The Final XCP Release also contains fixes to the following Xen security vulnerabilities. For more information on Xen security
vulnerabilities and fixes, see Xen Security Announcements.
- CVE-2012-4536 - XSA-21: pirq range check DoS vulnerability
- CVE-2012-4535 - XSA-20: Timer overflow DoS vulnerability
- CVE-2012-4537 - XSA-22: Memory mapping failure DoS vulnerability
- CVE-2012-4538 - XSA-23: Unhooking empty PAE entries DoS vulnerability
- CVE-2012-4539 - XSA-24: Grant table hypercall infinite loop DoS vulnerability
- CVE-2012-5457: Unprivileged XENMEM_add_to_physmap XENMAPSPACE_vlapic hypercalls can crash Xen
|XCP 1.6 Beta 4
Changes to previous build:
- Removed upgrading regression that was introduced in beta #3
|XCP 1.6 Beta 3
Changes to previous build:
- Allow upgrades from XCP 1.1 and from the previous XCP 1.6 betas
|XCP 1.6 Beta 2
Changes to previous build:
- Removed remaining XenServer branding
- Enable upgrading from XCP 1.5, and migrations from XCP 1.5 hosts during an upgrade
- Fixed performance issue in the event subsystem
|XCP 1.6 Beta 1
Known Issues and Errata
This section details known issues with this release and any workarounds that can be applied.
Installation and Upgrade
- RHEL, OEL, and CentOS 5.0 guest operating systems with the original kernel will fail to boot on XCP 1.6. Before attempting to upgrade a XCP host to version 1.6,
users should update the kernel to version 5.4 (2.6.18-164.el5xen) or later. Users running these guests who have already upgraded their host to XCP 1.6. [CA-79505]
- RHEL 4.5 guests may crash when when performing a rolling pool upgrade. Before upgrading a XCP host, you must shut
down RHEL 4.5 guests. [CA-88618]
- Rolling Pool Upgrade should not be used with Boot from SAN environments.
- When installing XCP from a network repository, you must configure the DHCP server to provide
the domain-name option, otherwise DNS will not work correctly, which can lead to a failed installation. [CA-74082]
- Shared storage should not be specified as the either the primary-disk or the guest-disk
during host installation. Storage specified during installation for both the primary-disk and the guest-disk will be formatted. [CA-41786]
- For Linux bridging users: During the upgrade to XCP 1.6 the network backend will be switched to Open vSwitch.
If you use an unsopported bondig mode (like LACP) and/or custom settings, Open vSwitch may not recognise these settings.
Revert any custom settings to your network interfaces (e.g. other-config settings for pifs) or ensure
your settings will work with Open vSwitch before the upgrade. Should you use an unsupported bondig mode (like LACP),
delete that bond before the upgrade and recreate it afterwards.
- Non-ASCII characters, such as characters with accents, cannot be used in the host console. [CA-40845]
- The root password of the host installer must not contain non-ASCII characters. [CA-47461]
- Intel Errata for Nehalem and Westmere CPU implementations of C-states may lead to system instability, apparently random freezes or reboots —
including during installation. [CA-60628]
- Brocade's BFA Fibre Channel over Ethernet (FCoE) driver version 3.1.0, as shipped with XCP 1.6, requires FCoE Initialization Protocol (FIP) login.
Users should ensure that the FCoE switch is correctly configured to accept such logins, otherwise access to the storage may be lost. [CA-88468]
- For active-active bonds on the vSwitch network stack, the bond rebalancing interval has changed from 10 seconds to 30 minutes. [CA-90457]
- When using Single Root I/O Virtualization (SR-IOV) enabled Intel NICs, the creation of Virtual Functions (VFs) has been disabled by default
in the driver. This will mean that any VMs with VFs passed through will refuse to boot. You may see an error such as
xenopsd internal error: Unix.Unix_error(19, "write", ""). To fix this, you must edit /etc/modprobe.d/igb (for Intel 1GBe NICs)
or /etc/modprobe.d/ixgbe (for Intel 10GBe NICs), and uncomment the line that looks like this: #options igb max_vfs=7,7,7,7 or
#options ixgbe max_vfs=40,40. [CA-89008].
- QoS settings do not work. [CA-90580]
- DHCP lease renewal fails if the DHCP client and DHCP server are both Linux VMs on the same host. This does not affect Windows VMs. If you
wish to use dhcp3-server and dhcp3-client in Linux VMs which may be sharing the same host, you must disable checksum offload by issuing
the command ethtool -K eth0 tx off within the DHCP server VM. [CA-40908]
- When using the vSwitch Controller with Microsoft Internet Explorer (IE) version 7 or 8 to access the vSwitch Controller remotely,
you may find that these versions of IE leak system resources. Citrix recommends using either Mozilla Firefox or IE 9, which addresses
some of the known IE memory and resource leak issues. [CA-65261]
- The vSwitch Controller may fail to show slave networks that had been bonded when NIC bonds are deleted. To resolve this issue, refresh
the status of the pool or restart the vSwitch Controller. The networks should then reappear. [CA-65261]
- XCP reports the amount of space used by a virtual disk (VDI), but this number may be substantially out of date. [CA-51466]
- When migrating VMs using Storage XenMotion, attempts to cancel the operation may not delete the temporary virtual disks. [CA-87710] [CA-87689]
- When a VM is unexpectedly shut down during a Storage XenMotion migration, the migration may fail. This process may leave a shut down
incomplete copy of the VM on the destination pool. Users should delete the VM from the destination pool and re try the operation with the VM running. [CA-86347]
- When using the xe CLI to migrate a VM with a snapshot using Storage XenMotion, you must provide a destination SR for each snapshot VDI.
This issue does not occur the first time that a VM is migrated. [CA-78901]
- Attempts to revert to a snapshot fail if the SR has too little space to inflate the snapshot. This is caused by a race condition and can be
resolved by re-attempting the revert after a few minutes. [CA-63032]
- Writing to CIFS ISO storage repositories is not supported and can result in disk corruption. [CA-41058]
- If an ISO SR is stored on an NFS server, and the connection to the server is temporarily lost, you may need to restart your XCP host in
order to regain connection. [CA-10471]
- If a Storage array reports IPv6 addresses the following error may be displayed: ValueError: too many values to unpack or
received signal: SIGSEGV. To workaround this issue, disable IPv6 on the storage array. [CA-90269] [CA-90271]
- Users using Dell EqualLogic arrays with Integrated Storage Link (iSL) should only use the firmware from the 5.x branch. This
firmware has been verified and tested for use with iSL.
- Users using Dell EqualLogic arrays with version 4.x or earlier firmware with Integrated StorageLink (iSL), may find Revert to
Snapshot operations fail. Users should upgrade their Dell array firmware to version 5.x. [CA-77976]
- Users using Dell EqualLogic arrays with Integrated StorageLink (iSL) performing certain manual delete options may encounter
the VDI not available error on attempts to reboot a VM based on array resources. [CA-78670]
- Users using EMC VNX series arrays with Integrated StorageLink (iSL) may experience issues when carrying out snapshot operations.
The array may hang and the iSL process time out. This incomplete operation may be incorrectly reported as having succeeded by iSL. In
#some cases, the array may recover automatically from this state and complete the snapshot task, creating a VDI which is unknown to the
XCP host. The workaround is to reboot the EMC VNX Storage Processor. [CA-90199]
- If a VM's VBD is unplugged and the VM is then rebooted, the VBD will remain unplugged after reboot. [CA-76612]
- Locking modes for VIFs may not be preserved when exporting and then re-importing a VM when using XVA or OVF formats. [CA-90857]
- The XCP SDK VM as shipped in previous version of XCP has been removed. Users should not attempt to use the Xen API
SDK VM template. As an alternative sandbox testing environment, you can install XCP as a generic HVM Guest using the
Other install media template (2048MB of memory and a disk size of at least 12GB is recommended). Note that the Guest's
IP address will not be reported though the CLI or XenCenter. [CA-89266]
- After using XenMotion (Live Migration) to move a Windows VM, the memory usage reported for the VM may be incorrect. [CA-89580]
- Verification of manifests and digital signatures on OVF and OVA packages will fail on import if the filename contains
parenthesis. The import will still succeed if verification is skipped. For the same reason, if you are exporting VMs as
an OVF/OVA package and are including a manifest or a digital signature, Citrix recommends specifying a package name
that does not contain parentheses. [CA-89555] [CA-90365]
- Attempts to detach a Virtual Disk Image (VDI) from a running a RHEL, CentOS, or OEL 6.1 and 6.2 (32-/64-bit) VM, may
be unsuccessful and can result in a guest kernel crash with a NULL pointer dereference at <xyz> error message. For more information, see Red Hat Bugzilla 773219. [CA-73512]
- A Windows VM may fail to boot correctly if streaming from PVS version 5.1. This is an intermittent fault, and rebooting
the VM should resolve the issue. [CA-60261]
- If you wish to create an Ubuntu 10.04 VM (32-bit) with more than 512MB of memory, you must upgrade to the latest
version of the kernel before increasing the RAM. For more information, see Ubuntu Launchpad 803811 and 790747. [CA-61400]
- Ubuntu 10.04 (64-bit) running the 2.6.32-32 #72 kernel, may crash with the following message, kernel BUG at
/build/build/linux-2.6.32/arch/x86/xen/spinlock.c:343!. The problem only affects VMs with multiple vCPUs.
vCPU hotplugging (only available via the xe CLI/API) should not be attempted with this guest. [CA-57168]
- Users running RHEL or CentOS 5.3 or 5.4 (32/64-bit) should not use Dynamic Memory Control (DMC) as this
may cause the guest to crash. If you wish to use DMC, Citrix recommends that users upgrade to more recent versions of RHEL or CentOS. [EXT-54]
- The RHEL 6.1 kernel has a bug which affects disk I/O on multiple virtualization platforms. This issue
causes VMs running RHEL 6.1 to lose interrupts. For more information, see Red Hat Bugzilla 681439, 603938 and 652262. [CA-60495]