[30353] [Salt] Changed the default value of the Glance image protected
type from string
to bool
to prevent the Glance state failures.
[29437] [Salt] Added the Cinder volume image_conversion_dir
configuration option to set the path to a directory
used for temporary storage during image conversion.
This prevents huge disk and input/output consumption on the controller nodes
if no volume nodes are used.
[30257] [Queens] Unhardcoded the RabbitMQ tuning settings in the Oslo templates Salt formula.
[30178] [Pike, Queens] Changed the Oslo messaging configuration to treat the SSL error timeouts as socket timeouts. This prevents the ConnectionForced: Too many heartbeats missed timeout errors on the compute nodes.
[29447] [Pike, Queens] Added the possibility to handle possible
TypeError
in on_inbound_method
in oslo_messaging
to raise
the Connection already closed. recoverable error
if another thread closes the RabbitMQ connection.
[29308] [Pike] Fixed the retry
logic in the RabbitMQ ensure
method to prevent the nova-compute
services from being unable to connect
to RabbitMQ after restart of the RabbitMQ cluster.
[30643] [Pike, Queens] Fixed the issue with Horizon displaying the
Unable to connect to Neutron: ‘frozenset’ object has no attribute
‘_getitem_’ error message in the Horizon logs causing the instances
list page to become empty. The error occurred during the network_list
requests when the parameter list was too long.
[29118] [Queens] Fixed the issue in Horizon that logged a user out with the keypair 403 (Quota Exceeded) error message from Nova. Previously, Horizon interpreted the Nova 403 error message as an unauthorized access and immediately logged the user out. Now, Horizon raises the corresponding Quota exceeded, too many key pairs. error message.
[29415] [Queens] Fixed the Horizon test failures in the test environments with Python 3.5 and Django 2.0.
[30447] [Salt] Unhardcoded the authentication URLs for the OpenID Connect (OIDC) protocol in the Keystone Salt formula.
[30027] [Salt] Set the etc/hosts
entries to lowercase
in the Keystone Salt formula to prevent the glance image-list failed.
HTTPMultipleChoices (HTTP 300) Requested version of OpenStack Images API
is not available. exception after an MCP cluster redeployment
with the cluster name containing capital letters.
[29843] [Pike] Fixed the issue in the OpenStack control plane upgrade from Pike to Queens of the OpenStack deployments with Octavia. Previously, the Deploy - upgrade control VMs pipeline job could fail with the NotFound: The resource could not be found. (HTTP 404) error.
[29844] [Pike] Fixed the issue in the
Deploy - upgrade control VMs pipeline job failing during the
application of the octavia.upgrade.verify._api state with
Internal Server Error (HTTP 500) due to a missing ca-cert.pem
on the MCP OpenStack deployments with Octavia.
[29862] [Pike, Queens] Fixed the issue with the dist-upgrade
stage
errors being ignored during the execution
of the OpenStack Deploy - upgrade control VMs
pipeline job.
The fix replaces runSaltProcessStep
in the osUpgrade
and
osDistUpgrade
functions with cmdRun
, which has the ability to verify
the return error codes.
[30149] [Pike, Queens] Fixed the issue with live migration of instances between the Pike and Queens-based compute nodes after all controller nodes are upgraded to Queens. In Queens, the controller nodes send events about NICs being plugged only to the source compute nodes, while the Pike compute nodes logic expects these events on the destination nodes. The fix passes these events to both source and destination compute nodes.
[29798] [Pike, Queens] Fixed the issue with the Barbican verification
failure during the update of the OpenStack control plane by adding retries
to the barbicanv1
client.
[29451] [Reclass] Fixed the issue with IP being used instead of FQDN in
OS_AUTH_URL
for keystonercv3
, keystonerc
files,
as well as in catalog and endpoint lists for admin and internal endpoints.
The fix sets FQDN instead of IP for OS_AUTH_URL
in keystonercv3
.
[29377] [Queens, Salt] Fixed the issue with the bootstrap procedure creating the Keystone admin user and executing on all OpenStack controller nodes. Added the cluster role node check to the Keystone Salt formula to execute the bootstrap procedure only once on the first OpenStack controller node only.
[29357] [Salt] Updated the port configuration for RabbitMQ in the Barbican metadata to consistently use the port 5672.
[29126] [Salt] Fixed the issue with the gnocchi.server
state failing
during an OpenStack cluster deployment by updating the Redis configuration
for all Ceilometer-related components.
[28990] [Salt] Fixed the potential security issue in an image corruption,
compromising, or being overwritten due to the Glance
show_multiple_locations
and show_image_direct_url
parameters being
set to True
by default. The fix sets these parameters to False
by default.
[28973] [Pike] Fixed the issue with inability to change the language to Chinese in Horizon dashboard.
[28896] [Pike Queens] Added support for the |
symbol to the VNC UI
in Horizon. To apply the fix on existing OpenStack environments, stop
the required instances using Nova or virsh and start them using Nova.
[28726] [Queens] Fixed the issue with the Horizon Angular-based
dashboards, for example, for uploading an image or creating a snapshot,
not being refreshed after an item on the dashboard page changes its status
from the transition state. For example, an image could remain in the endless
Creating
state until the page is reloaded. The fix adds a periodical
check for the status of an item being in the transition state.
[28711] [Pike Queens] Fixed the issue with the discrepancy between
the availability zone name of an aggregate and a host
after renaming the availability zone for an aggregate. Now,
PUT /os-aggregates/{aggregate_id}
and
POST /os-aggregates/{aggregate_id}/action
return an HTTP 400 message
during an availability zone renaming
if the hosts of an aggregate have any instances.
[28688] [Pike] Fixed the following issues with image upload failures when Glance is used with the Swift back end:
Fixed the issue with renewing connections to Swift by adding usage
of the cached auth_ref
property instead of obtaining a new one
each time a given token is about to expire.
Fixed the issue with the last chunk being left in storage when an image upload to the Swift back end fails. Previously, chunks were counted after uploading, now, they are counted before uploading.
[28616] [Queens] Fixed the issue with glance-api
failing to validate
self-signed certificates when using Glance with Swift back end
and SSL enabled.
[27017] [Salt] Fixed the issue with inability to configure the Heat
reauthentication methods by adding the possibility to set
the reauthentication_auth_method
parameter for Heat
in the Reclass cluster model.
[29603] [Salt] Fixed the issue with loss of connectivity to a guest VM
after applying configuration changes to the Neutron OVS bridges.
The fix adds linux:network:libvirt_vnet_repair: true
to automatically
reattach the libvirt vnet
interfaces by getting the information
about them from the current libvirt definitions to repair a guest VM
connectivity with the host.
[27276] [Salt] Fixed the issue with importing errors and
loading warnings on the Salt Master node from the neutronv2
modules.
[30150] [Pike] Fixed the issue with the MAC address is in use error when migrating a VM with a direct-physical port if the SR-IOV Physical Function (PF) passthrough (PT) ports are used. The fix adds a reset of a MAC address when unbinding a direct-physical port from a VM.
[29402] [Salt] Fixed the issue with the Neutron Salt formula being unable to handle the default quotas by adding a capability to configure them.
[29110] [Pike] Fixed the issue with failures during bulk creation of IPv6 subnets using API, leading to HTTP/1.1 500 Internal Server Error.
[29040] [Salt] Added the capability to change the hardcoded values in the
Cinder, Glance, and Nova configuration files by adding the
configmap
pillars to related Salt formulas.
[25928] [Pike] Fixed the issue that caused approximately 5% of instances to fail during the live migration of QEMU v2.5 to v2.11.
[30410] [Queens] Fixed the issue with inability to create a new VM after upgrading an MCP cluster.
[27222] Fixed the issue with the /var/log/glance/api.log
permissions causing the upgrade of OpenStack to fail with the Salt state on
controller nodes (``ctl*``) failed exception.
[31278] [Pike] Fixed the issue with MySQL occasionally failing on the Galera nodes.