[Xen-users] RE: [Xen-API] [XCP]: RC1 of XCP 0.5 available for testing

[Dave Scott <Dave.Scott@xxxxxxxxxxxxx>]

Hi Marco,

XCP 0.5 is primarily intended to be a bugfix release. However it does have 2 
new features:

1. CPU feature masking (via Intel FlexMigration and AMD-V Extended Migration)
   -- this allows you to (eg) "down-level" a new server to make it look the 
same as an old server. The machines can then be joined to the same pool and you 
can be sure that live migrations will work fine.

2. the ability to create and upload ISO images (provided the underlying SR is 
itself writable)

Here's a short summary of the bugfixes, per area, since 0.1.1:

Toolstack:

fix a race in the Host_patch record update logic
Revert c/s 594:41501bb83b18 (session caching stuff)
Fix XenAPI performance degradation caused by RBAC
Support for likewise 5.3 error codes
Add first-class field for openvswitch distributed controller
Fix PIF_metrics glitch when xapi is restarted
Fix ballooning bug where a domain's maxmem gets out of sync in the no 
ballooning required case (ie dynamic_min = dynamic_max)
Fix terrible error message when a 'rhlike' guest (eg CentOS) cannot be 
installed because of a missing file
Read-only RBAC roles can call 
pool.ha_compute_hypothetical_max_failures_to_tolerate
vswitch: the MAX address of the tagged PIF for VLAN fake-bridge
any PAM user should be able to log-in when external auth is disabled
robustify VBD unpause in the case when the device disappears
Make sure we can't VBD.plug while lvhd_stop_using_these_vdis is in progress
Make it possible to use new memory APIs on templates as well as VMs
Always allow dom0 to be ballooned even if other domUs cannot be ballooned
Resync Host_cpu tables with physical h/w on restart
Fix memory glitch when a domain is paused (eg while being resumed)
Fix a deadlock in the RRD code
Fix a glitch where xapi would set PV-drivers-up-to-date to false on restart
Extend VM.clone and VM.copy to work on snapshots and checkpoints
Add "Xen API SDK" template: using this to install from the main CD should 
result in a simulator
Work around broken vendor kernels by forcing all PV guests to use slow resume 
during checkpoint operations
Fix periodic backup of RRD data: broken by fork/exec daemon
Fix occasional HTTP request failure caused by failing to read all request 
headers from an HTTP PUT
Improve the RRD behaviour when the system clock moves backwards
When importing VM metadata, if the VM was Running then the new state should be 
Halted (not Running)
vswitch: fix bringing up VLAN PIFs
When deciding which host to forward a VDI.copy to (eg when installing a VM to a 
different SR), take into account host liveness
SRIOV: on reboot, always hotplug PCI devices into guests in the order specified 
in the (original) other-config key
Fix bug where 'xe pif-scan' could pick up non-physical interfaces such as vif 
backends (!)
Avoid leaking a domain in the case where a storage backend fails during resume
Allow VCPU settings to be changed on a template
Allow the CLI to change a Host's name-description
Work around firewalls which kill idle TCP connections by inserting empty blocks 
every 5s or so
Fix segfault in fork/exec daemon
Accept ISO8601 dates without a timezone, assume UTC (XMLRPC spec problems)

Dom0 kernel:

Fix deadlock in software raid by importing upstream fixes.
Enable CONFIG_NET_SCH_INGRESS
ensure gratuitous frames sent on bond link failure are truly benign
Allow large MTU to be configured on netback vif device from hotplug script.
pciback: fixes for SR-IOV passthrough to PV guests.
Fix iSCSI last_recv timestamp hijacking in iscsi_xmit.
Prevent iSCSI transport timeout checks from spinning.
Refresh Intel NIC drivers from http://e1000.sf.net
Do not restrict netback to the first 2k of each RX buffer.
Backport xenbus_watch_pathfmt which is used in place of xenbus_watch_path2 in 
pvops kernels.
compile pciback in the kernel (not as a module)

xen (3.4.2):

EPT/PoD locking fix 
Backport upstream patch to support syslinux configuration files in pygrub.
add AES to the HVM CPUID whitelist.
Import upstream patch to improve pygrub compatibility with grub2.
Upgrade gPXE to 1.0.0
don't complain too loudly about mixed-CPU migrations.
new bypercall to get the unmasked CPUID feature bits
Mask CPUID masks with unmasked CPUID responses.
drop CRC bytes from RX frames in the gPXE rtl8139 driver.
hvm: correct time offset update in RTC write emulation
Import xentrace patches from unstable, retry
pygrub: Fix grub2 support when config is on a separate /boot partit
Added patch to fix network booting from Bootix serverion
quieten per-cpu logging and increase dmesg buffer size.
import Intel changes to FlexMigration
trace possible incorrect use of SPIN_LOCK_UNLOCKED
With OOS enabled, check l1 and rewalk if it's changed just before propagation.
improve support for grub2 and reiserfs/ext4
CA-40782: Sort out interactions between PoD and log-dirty on EPT.
pygrub: Fix Grub2 support for Ubuntu 10.04

storage plugins

Fix SAS dynamic device removal/addition behaviour
Fix race in waiting for LUNs to appear. 
if we fail to remove a devmapper node, leave the LV symlink
retry on failures during "dmsetup remove" (since most of those are transient)
Run NFS slave callout before unlinking parent VHD.
Activate the dm-multipath node before calling vdi_attach
enable single snapshots (1 leaf instead of 2), which would have been needed for 
online leaf-coalesce
Make VDI.introduce on LUN-per-VDI backends set managed=true on the  VDI.
Add Multi-IQN wildcard support to the iSCSI target discovery functionality.
fix a journal file race
use the BAT information to minimize the space requirements of coalesce
fix the space requirement calculation in vdi_snapshot
make vhd-util log errors to syslog (except for vhd-util scan).
log some diagnostic info when LV fails to get activated
fix the clone rollback call params
Ensure that root dev does not get re-probed when refreshing the SCSI host bus.
Report the correct number of iSCSI sessions when the MultiIQN wildcard is being 
used.
don't trust LVM when it says a LV doesn't exist: still check if the devmapper 
node and/or symlink do exist and delete them manually
Restore set_scheduler device name for multipathed nodes.
VM template should only be created when snapshotting via 
vm-snapshot-with-quiesce.
zero out the footer area of the LV when rolling back inflates

Cheers,
Dave


> -----Original Message-----
> From: Marco Sinhoreli [mailto:msinhore@xxxxxxxxx]
> Sent: 12 June 2010 22:13
> To: Dave Scott
> Cc: xen-api@xxxxxxxxxxxxxxxxxxx; xen-devel@xxxxxxxxxxxxxxxxxxx; xen-
> users@xxxxxxxxxxxxxxxxxxx
> Subject: Re: [Xen-API] [XCP]: RC1 of XCP 0.5 available for testing
> 
> Hi Dave,
> 
> What are the features and the bugfixes in this new release?
> 
> Cheers!
> 
> On Fri, Jun 11, 2010 at 5:27 PM, Dave Scott <Dave.Scott@xxxxxxxxxxxxx>
> wrote:
> > Hi everyone,
> >
> > The first release candidate of the Xen Cloud Platform (XCP) version
> 0.5 is
> > now available for testing from:
> >
> > http://www.xen.org/products/cloud_source_0.5.html
> >
> > XCP-0.5 is intended to be a *stable* release, suitable for long-term
> production use.
> > Please download this release candidate and give it a thorough workout!
> >
> > Cheers,
> > Dave
> >
> > _______________________________________________
> > xen-api mailing list
> > xen-api@xxxxxxxxxxxxxxxxxxx
> > http://lists.xensource.com/mailman/listinfo/xen-api
> >
> 
> 
> 
> --
> Marco Sinhoreli

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users