RHBA-2014:0548 - Bug Fix Advisory

Topic

Updated vdsm packages that fix a bug are now available.

Description

VDSM is a management module that serves as a Red Hat Enterprise Virtualization
Manager agent on Red Hat Enterprise Virtualization Hypervisor or Red Hat
Enterprise Linux hosts.

All users managing Red Hat Enterprise Linux Virtualization hosts using Red Hat
Enterprise Virtualization Manager are advised to install these updated packages,
which fixes these issues.

Changes to the vdsm component:

• Previously, migration start time was captured at the start of the

MigrationSourceThread process. This meant that the migration would fail if the
virtual machine had to wait a long time to acquire the migration semaphore. Now,
the migration start time is captured when migration begins. (BZ#1097341)

• Previously, VDSM would not start correctly under certain conditions after

rebooting a hypervisor. This was caused by qemu-sanlock.conf not being persisted
after rebooting the hypervisor when the command 'vdsm-tool configure --force'
was run and libvirt sanlock was enabled. Now, qemu-sanlock.conf is persisted
even after rebooting hypervisors, making it possible for VDSM to start
correctly. (BZ#1095303)

• * Red Hat Enterprise Virtualization Hypervisor 6.5 added support for

memory-based file systems (for the purposes of testing), which included code
with O(N^2) complexity. This introduced performance issues, such as high cpu
consumption by vdsm, in certain environments that utilized numerous storage
domains. Now, the additional support for memory-based file systems has been
disabled and does not contribute to performance problems. (BZ#1095907)

• Previously, VDSM would report that virtual machines experiencing any I/O error

were in a paused state. This was caused by the logic used by VDSM to check I/O
errors received from libvirt. Now, the logic used to check such errors has been
revised so that VDSM detects the nature of the error, allowing I/O errors to be
correctly reported and handled. (BZ#1090079)

• Previously, migration of virtual machines between hypervisors would fail under

certain conditions, causing the virtual machine being migrated to crash. This
was caused by the logic used to clean up sockets for virtual machines after VDSM
restarted due to an unexpected failure. Now, the logic used to clean up sockets
for virtual machines has been revised so that VDSM correctly cleans up sockets
after VDSM is restarted, allowing virtual machines to migrate without issue.
(BZ#1088906)

• Previously, the getStorageDomainInfo action of the vdsClient command would

fail when run under certain conditions. Now, the logic used to retrieve
information on storage domains using the vdsClient command has been revised so
that the action returns information correctly. (BZ#1084970)

• Previously, the vmHotplugDisk action would fail with the error "VolumeError:

Bad volume specification". This was caused by a multipath race condition that
would result in a situation in which multipath was not given enough time to
create the /dev/mapper entry for the LUN on the host on which the virtual
machine was running. Now, the logic used to scan for device visibility has been
revised so that multipath has the time required for creating the /dev/mapper
entry for the LUN on the host on which the virtual machine is running.
(BZ#1066413)

• Previously, VDSM would report that a storage domain was active even although a

link for that storage domain was not present in the /rhev/data-center/SPUUID
directory. This would occur when the host was in a non-operational state and was
able to connect to the storage pool but was not able to connect to any storage
domains. Now, the logic for refreshing storage domain links has been revised so
that those links will be created correctly for storage domains even under these
circumstances. (BZ#1066409)

Solution

Before applying this update, make sure all previously-released errata
relevant to your system have been applied.

This update is available via the Red Hat Network. Details on how to
use the Red Hat Network to apply this update are available at
https://access.redhat.com/site/articles/11258

Affected Products

• Red Hat Virtualization 3.3 x86_64
• Red Hat Virtualization 3 for RHEL 6 x86_64

Fixes

• BZ - 1066409 - [vdsm] storage domain is reported as 'active' by host, even though its link under /rhev/data-center/SPUUID/ is missing
• BZ - 1066413 - vmHotplugDisk failed with "VolumeError: Bad volume specification"
• BZ - 1084946 - vdsm libvirt and sanlock daemons need to be restarted after upgrading to sanlock 2.8
• BZ - 1084970 - [vdsm] getStorageDomainInfo fails due to key 'info' missing from poolInfo for master domain
• BZ - 1088906 - VM migration back to original host fail on "file exists"
• BZ - 1090079 - vdsm reports guest as paused on any IO error, even if libvirt/qemu policy is set to "report"
• BZ - 1095303 - /etc/libvirt/qemu-sanlock is not persisted and vdsm fails to start on reboot
• BZ - 1095907 - VDSM is consuming a lot of cpu time even with no active VMs
• BZ - 1097341 - The start time for 'migration_max_time_per_gib_mem' appears to be calculated too early.

CVEs

(none)

References

(none)