Add dcorch-identity-api-proxy ocf script as part of Distributed Keystone
for Distributed Cloud feature.
Story: 2002842
Task: 22785
Change-Id: Icd215916e27db785af145c68c72d13abdaf577ba
Signed-off-by: Jack Ding <jack.ding@windriver.com>
Changes included in the patch:
- Add missing init files
- Update MANIFEST.in to include non-python files
- Add the ability to launch gnocchi api with gunicorn
Story: 2002825
Task: 22871
Change-Id: Iad1ed6e4d6cd4c4debfd8ed7845d4de69af3ea14
Signed-off-by: Jack Ding <jack.ding@windriver.com>
- Support new gnocchiclient interface in openstack-aodh
Ported fix from openstack-aodh 5.0.0.
- Update heat templates to create gnocchi resources type of alarm
instead of ceilometer
- Remove all the heat templates using cfn-push-stats tool to push
metrics from instance, as heat CloudWatch API/cfn-push-stats is
deprecated for a long time and it was removed in openstack queens.
references:http://lists.openstack.org/pipermail/openstack-dev/2017-October/123104.htmlhttps://bugs.launchpad.net/heat/+bug/1743707
- updates the scaling interval to 5 minutes in heat templates as the
granularity of the archive policy we are using in gnocchi is 5
minutes.
Story: 2002825
Task: 22871
Depends-On: https://review.openstack.org/587279
Change-Id: I4872c67d15065c0e7a16632390488305649f8f37
Signed-off-by: Don Penney <don.penney@windriver.com>
Signed-off-by: Jack Ding <jack.ding@windriver.com>
Add python-gnocchi and python-gnocchiclient.
Update in python-openstackclient:
Remove the support of celiometer CLI extensions.
Update in python-ceilometer:
Install ceilometer publisher instead of dispatcher.
Story: 2002825
Task: 22871
Depends-On: https://review.openstack.org/587279
Change-Id: I3b0dde2c8668f7e623bcf128a13010b26667d802
Signed-off-by: Don Penney <don.penney@windriver.com>
Signed-off-by: Jack Ding <jack.ding@windriver.com>
In Distributed Cloud, Keystone is now running on each Subcloud.
Switching to Subcloud region now requires Openstack Auth to retrieve an
Unscoped token from the switched Region and reinitialize the django
session and cookie data with token data retrieved from the Subcloud.
Since Subcloud's Keystone doesn't contain the Identity endpoint for the
Central Region, there was no way to go back in Horizon from a subcloud
region to the SystemController region. We achieve this by caching the
SystemController endpoint in the Django Session at the time of login.
Story: 2002842
Task: 22785
Change-Id: I274e8b0756e0f24321a108c6c1a0a5d6178e0c7a
Signed-off-by: Jack Ding <jack.ding@windriver.com>
This is part of Distributed Keystone for Distributed Cloud feature.
Story: 2002842
Task: 22785
Change-Id: Ia8d2e3371e74ce8994a8c23ff926e5b6816b4f51
Signed-off-by: Jack Ding <jack.ding@windriver.com>
This is part of Distributed Cloud keystone scalability.
Signed-off-by: Andy Ning <andy.ning@windriver.com>
Change-Id: Ia39bf0b7a53d28188bb0cf9585d0b15a0ea66539
Story: 2002842
Task: 22785
Signed-off-by: Don Penney <don.penney@windriver.com>
Signed-off-by: Jack Ding <jack.ding@windriver.com>
Fixed glance uploads on poor write performing systems: give other
greenthreads a chance to schedule.
Story: 2002893
Task: 22856
Change-Id: I70c192f1c72d8bb1a9f9b7a0fd2545c49c3acb46
Signed-off-by: Jack Ding <jack.ding@windriver.com>
Ceph Cache Tiering feature is not supported anymore. This commit removes
all the code changes associated with the no longer supported Ceph Cache
Tiering.
This implies:
a. cache tiering cannot be configured on system
b. no ceph-caching host could be added
c. no ceph-backing host could be added
d. ceph-caching/ceph-backing personality sub-type won't show up
when 'system host-show'/system host-add command is issued
e. ceph-caching/ceph-backing personality sub-type won't show up when host
is added/listed from horizon
Change-Id: Idfe9a1672d17ecb996069311e2ad9377921b3df0
Signed-off-by: Jack Ding <jack.ding@windriver.com>
TIS_WORKERS_FACTOR for keystone service is reduced to 1 for all AIO
variants.
Tests conducted:
- Bulk VM launch/delete via heat stack create/delete - duplex (20 VMs) &
simplex (10 VMs)
- Swact, failover, parallel live migration - duplex
- Lock/unlock with max VMs, DOR - both duplex & simplex
- Sanity, nightly sanity - both duplex & simplex
- Full nova regression - duplex
- Bulk VM launch/delete stress test - duplex (60 VMs)
- Platform CPU adjustments via system host-cpu-modify
- Ceilometer message queues verification
- Standard system installation
Change-Id: Ia455f1492ed3950ec937ad6d3a2e08b2364d9665
Signed-off-by: Jack Ding <jack.ding@windriver.com>
A timeout between ceph-manager and sysinv is causing ceph-manager to
stop responding. When Ceph Manager detects that 'require_jewel_osds'
flag needs to be set it queries sysinv which for whatever reason
(most likely dealing with some final operations) fails to respond in
a reasonable amount of time. This causes an exception in ceph-manager
which breaks one of the periodic threads execution.
Change-Id: Iac5a4cd171cd702d84053fba7860bea244b31f4e
Signed-off-by: Kristine Bujold <kristine.bujold@windriver.com>
- Reduce both CPU and memory footprint for AIO simplex by reducing the number of workers across the board.
- In addition:
a. Reduce ceilometer workers to 1 as the traffic generated from max number of VMs in simplex (20) does
not warrant more workers than 1. Increase batch timeout to reduce load.
b. Change keystone workers factor from 1.5 to 1 which further reduces the number of keystone workers
for simplex. The number of requests do not warrrant more workers than other services.
c. Reduce the number of nova api metadata workers to half of nova api compute workers for both AIO and
standard configs based on LO data.
AIO installations, simplex sanity & sanity nightly all passed. Sanity, parallel VM/volume launch timings
also improved. Ceilometer messages were cleared timely.
Change-Id: I86455e64b48fdc543cd9e99cf95f5ddb2295f71f
Signed-off-by: Scott Little <scott.little@windriver.com>
Bringing in the Endpoint Filter Group blueprint into Openstack
client. Endpoint filter groups have been around in Keystone since the
days of KILO but only picked up by Openstack client in v3.15.
In the absence of Endpoint filter groups, clients are responsible for
filtering the endpoint list, instead of Keystone sending a tailored
list. For e.g. in Distributed Cloud, local Openstack services in the
Central Region should not be seeing endpoints for subclouds, as they
don't directly communicate with subclouds.
As part of this blueprint we also had to bring in an important
commit that Optimizes the command "openstack endpoint list". The current
behavior is to make a Service call for each Endpoint in the Endpoint
list, so for N endpoints yo have N + 1 calls to Keystone. The new
behavior makes 1 call for Endpoint list, and 1 call for Service list and
works with the two lists... so you do down from (N+1) calls to 2
Change-Id: I9c03938c25b56d64b59ce42cae5026f2830f02b7
Signed-off-by: Tyler Smith <tyler.smith@windriver.com>
Problem:
The centos-pkg-dirs files should only reference packages with
compilation instructions hosted in the same git.
Solution:
Create centos-pkg-dirs files in other stx-* gits, and relocate
the relevant entries from the stx-utils centos-pkg-dirs into
the appropriate destination git.
Change-Id: I31c38ced2227aa494060c0ce333b569a64cc0424
Signed-off-by: Scott Little <scott.little@windriver.com>
Integrates the latest Open vSwitch with DPDK into the host management
and configuration framework and configures the default system
vswitch type to be ovs-dpdk.
Change-Id: I943f7668d9c94508706fa1611444898fc28ad4c7
Signed-off-by: Matt Peters <matt.peters@windriver.com>
