As of 0.12.14, fluentbit exposes a flag for setting the db_sync
behavior for writing the location of the tail input to its sqlite
database. The default setting is Full, which introduces additional
synchronizations before and after a transaction. This has the
potential to negatively affect disk performance with the extra
synchronizations. This moves the setting in the chart to Normal,
which performs fewer synchronizations and still maintains a high
level of safety with status writes
Change-Id: I5de8ee40162a50e1ff3395d0e302974b77dd5f9c
This PS remove change pvc-tempest.yaml from manifests.pvc to
pvc.enabled as there is no manifests enabled value for pvc.
Bug was tempest pod is in pending state as pvc is not found.
Change-Id: Ib6704725c7415e95237f633d340479e33b43d85b
This adds support for executing helm tests via the armada test
directive. It enables theses tests for all services, except for
nova and neutron as executing tests with armada force a chart to
wait. Forcing nova and neutron to wait effectively sequences the
charts, which will result in a failure to deploy past those
services
Depends-On: https://review.openstack.org/#/c/581148
Change-Id: I6ac845c82d744e2f5fd79c3e2ff3c1479dd1ddab
nova-console-compute-init.sh and nove_consolt-proxy-init.sh generates
incorrect configuration if there exists multiple IP address on default
interface.
To solve this problem, we pickup first IP address if there exists
multiple IP on that interface.
Change-Id: Iaadd2e71d624122e68fdd628771df21cd61c0784
This PS updates the ceph charts to use endpoints for all port
lookups, and also prunes some unused defaults.
Change-Id: I14c8eff3d4d6290dd98e90029abc21a55713d4e5
Signed-off-by: Pete Birley <pete@port.direct>
This reverts commit 671166e064.
As it has no impact on the ingress controller we use, and was merged in error.
Change-Id: I9e3a23ffa0e9af8bef9ef54062e55f635e4f376d
This PS updates the RGW config to not cache keystone tokens to
work round expiry bug: http://tracker.ceph.com/issues/21226
Change-Id: Ice80c2510085215f03ce0977c8a4bff9ddb2f07a
Signed-off-by: Pete Birley <pete@port.direct>
While YAML allow 4 spaces to define a dict, being consistent
with two spaces provides clarity for the deployer.
Change-Id: Iff2aa15267205a04b58bc0324af5eb26ad937f52
This PS splits out the ceph provisioners from the client chart
allowing OSH to be used with external (and multiple) cephs more
simply.
Change-Id: I0c5d516c7a9a10bc8435e2423eb1dc05460d2c92
Signed-off-by: Pete Birley <pete@port.direct>
This patch sets the notifier driver to messagingv2 so all
messages sent to RabbitMQ using the 2.0 format.
For info:
Auth notificaitons are supressed by default, to enable it
uncomment the following parameter:
conf:
DEFAULT:
notification_opt_out: ""
Change-Id: I652fc34a229612dcb8b96c8722b8c6ac1c8aba3e
Signed-off-by: Ruslan Khanbikov <rk760n@att.com>
This PS updates the agent init container scripts to ensure that the
local IP address is populated on agent startup.
Change-Id: Ib5857d9dc82b1138a1b5f496ebe46dc1e3f221e7
Signed-off-by: Pete Birley <pete@port.direct>
This PS updates the glance management jobs to fail on incorrect
configuration, or abnormal interaction between components.
Change-Id: I87eb8f28bfaa93467f2690d8a7f92b062f9d257f
Signed-off-by: Pete Birley <pete@port.direct>
The current rally job is always execute report command at the end,
and it always succeeds regardless of the rally task result.
Since the values.yaml in the rally chart already includes sla check,
you should execute sla-check command at the end of the job
so that you can check whether the test has passed or not.
Change-Id: I4c91950b86622b1211aea3e9e908cb04711c0e76
This PS updates the armada manifests to delete all jobs for a release
via the `release_group` label. It also extends this to cover charts that
do not currently have jobs contained, to allow simple future expansion
of services.
Change-Id: I28bd9e2c2ebc22ca9c4319970c657298281e28d2
Signed-off-by: Pete Birley <pete@port.direct>
This PS makes some minor updates (links, formatting)
to the Multinode guide and adds some helpful warnings
and notes.
Story: 2002526
Task: 22064
Change-Id: I21a50b7126c2bb1fcd412a8ae713856e9fab4050
We should replace role name from deploy-yq to deploy-jq.
Depends-On: https://review.openstack.org/578703
Change-Id: I5deb4fecbee9f70e6be63189d4afa81e5c71ad4b
Keystone token expiration, rotation frequency and active keys
values should follow the formula:
max_active_keys = (token_expiration / rotation_frequency) + 2
max_active_keys by default is 3
token expiration set to 43200 (12h)
rotation frequency set to 12 hours
Change-Id: Ia04daec9b2905ef2d3f2d4fbb43557dda220dc70
Signed-off-by: Ruslan Khanbikov <rk760n@att.com>
This PS updates the Nova and Libvirt charts to mount the vm state
directories with bidirectional mount propagation for k8s >= 1.10.
This allows mounts created by some volume drivers to be used, and
unaffected by pod restart.
Change-Id: Idaf664efb23a424dd8d9e1376ea7231b8565e3fe
Signed-off-by: Pete Birley <pete@port.direct>
Ideally, this would completely own the postgresql.conf file that
is autogenerated by the container. This should not be done without
further deliberation though on what impact that may have. For now,
this allows some critical items to be manipulated at installation
time but does not meet the full need of complete configuration
flexibility.
Change-Id: I94fd3ac42398f8114c0b5c8b4ef0a6a96e530f96
This PS moves the neutron agents to run as child processes of either
the pause container or use the hosts init system (for k8s <1.10)
to prevent defunct process sprawl.
Change-Id: I3392bdc957144c1aa83314583d57183d35279336
Co-Authored-By: Hyunsun Moon <hyunsun.moon@gmail.com>
Signed-off-by: Pete Birley <pete@port.direct>
This PS moves to use the swift backend for glance by default, which
in the case of OSH is currently served by keystone auth'd radosgw.
This change moves the chart to be inline with the current gates, and
deployments - which have been using swift by default for some time.
Change-Id: Ia9c954ae2bd833e7f449bfdf7c51f8df5c78ba57
Signed-off-by: Pete Birley <pete@port.direct>
This introduces a mechanism for generating the logging.conf
file for the openstack services via the values. This allows us to
define loggers, handlers, and formatters for the services and the
modules they're composed of.
This also allows us to take advantage of the oslo fluent handler
and formatter. The fluent handler and formatter give us the
following benefits: sending logs directly to fluentd instead of
routed to stdout/stderr and then through fluentbit to fluentd,
project specific tags on the logged events (enables us to define
more robust filters in fluentd for aggregation if required),
full traceback support, and additional metadata (modules that
created logged event, etc)
Depends-On: https://review.openstack.org/577796
Change-Id: I63340ce6b03191d93a74d9ac6947f0b49b8a1a39
