This change adds feature to launch Prometheus process using a custom script which should be stored in override values. Because the known issue https://github.com/prometheus/prometheus/issues/6934 is still open many years, we are going to struggle with growing WAL files using our custom downstream wrapper script which stops Prometheus process and deletes WALs.
This solution can not fit all customers because completely kills wal cached data but it is ok for our purposes. Such way I just added the feature to use another custom script to launch Prometheus and left original functionality by default. Default/custom mode are defined in 'values.yaml' as the body of the custom launcher script.
Change-Id: Ie02ea1d6a7de5c676e2e96f3dcd6aca172af4afb
This change adjusts the minimum OSD count check to be based on the
osd value, and the maxiumum OSD count check to be based on the
final_osd value. This logic supports both full deployments and
partial deployments, with the caveat that it may allow partial
deployments to over-provision storage.
Change-Id: I93aac65df850e686f92347d406cd5bb5a803659d
The target OSD count and the final target OSD count may differ in
cases where a deployment may not include all of the hardware it is
expected to include eventually. This change corrects the check for
more OSDs running than expected to be based on the final OSD count
rather than the intermediate one to avoid false failures when the
intermediate target is exceeded and the final target is not.
Change-Id: I03a13cfe3b9053b6abc5d961426e7a8e92743808
tmp is an emptyDir volume mount with 0777 permissions. Some versions
of Ruby require the sticky bit to be set.
Change-Id: Ib6f4daa2068e7b29c62b5858848774b7117f6808
osh-infra still have references to fedora 27 which is quite old now,
this change removes those definitions since they are unused.
Change-Id: I4f127113f0014ec2ed11f21e230facd08820af6e
This change modifies the versions of both minikube
and kubernetes that we deploy for OSH.
minikube is now 1.25.2
kubernetes is 1.23.12
Change-Id: Ic33c1b1ca0110bdd87705ca5b891823f16b303f8
A couple of the ingress images are still pointing to the older
stein release of neutron. This change updates them to use the
updated xena release.
Change-Id: I95aecec5474e587d01d7e8812ec662fbf46ca634
The Ceph Pacific release has added a noautoscale flag to enable
and disable the PG autoscaler for all pools globally. This change
utilizes this flag for enabling and disabling autoscaler when the
Ceph major version is greater than or equal to 16.
Change-Id: Iaa3f2d238850eb413f26b82d75b5f6835980877f
Updated postgres binary version to 14.5.
Also replaced deprecated config item wal_keep_segments with wal_keep_size.
Change-Id: Ie86850f8ebb8bfaae4ba5457409d3920b230ce9c
HTK - added verify_databases_backup_in_directory function that is
going to be defined inside mariadb/postgresql/etcd charts.
Mariadb chart - added verify_databases_backup_archives function
implementation.
Added mariadb-verify container to mariadb-backup cronjob to run
verification process.
Added remove backup verification pocess - comparition of local and remote file md5 hashes.
PostgreSQL chart - added empty implementation of verify_databases_backup_archives() function. This is a subject for future realization.
Change-Id: I361cdb92c66b0b27539997d697adfd1e93c9a29d
This PS updates Grafana to 8.5.10 and enables unified alerting
which will be the default in 9.x.
The following directories are required for unified alerting:
/var/lib/grafana/alerting
/var/lib/grafana/csv
If a user is upgrading from Grafana 7.x and wants to opt out of unified
alerting they will need to set:
[alerting]
enabled = true
[unified_alerting]
enabled = false
in grafana.ini. Additionally, to roll back, the env var
GF_DEFAULT_FORCE_MIGRATION: true
needs to be set for the grafana_run_migrator pod.
Unified alerting doc:
https://grafana.com/docs/grafana/v9.0/alerting/migrating-alerts/
Change-Id: I3a6ca005b9d9433e958802e7e978b81479a16fb8
In an environment with helmv3, it was noticed that the mariadb
helmrelease is failing to render properly due to unsupported map key
type (int).
This change quickly fix this problem by quoting the value, forcing it to
be rendered as a string.
Signed-off-by: Thales Elero Cervi <thaleselero.cervi@windriver.com>
Change-Id: I2f2be87d0f79ca439e731d07354bcd5f149790d5
When we use Calico v3.23, docs.projectcalico.org is
redirected to projectcalico.docs.tigera.io and moved manifests
to folder archive.
Calico v3.20 present in both locations.
Change-Id: I653fbac92c4ec5c2a53670391658a50a25fe81a0
At the moment, our publishing scripts have their own little
way of deploying Helm so it's using the old version of Helm
which is failing linting.
This updates it so that it matches how we're running it inside
our lint playbook, and it will also fix the builds since it's
also failing because of the old version of Helm.
Change-Id: I719c2dadc3ca87912234ac13d87d63e8c7b779a7
We broke the chart builds when we dropped the 'package' target in the
Makefile. This fixes it also removes/drops the need to run the sed
locally since we point to the local folders anyways.
Change-Id: Iac8924f14c8fbb8e07b96b51790539c3ce1618e6
This change updates the kibana indices creation to repeatedly make
call attempts until we get a 200 response back.
Change-Id: Id0f012bda83913fc66c4ce105de97496043e487c
Having the "use_external_ingress_controller" field in
"network.server.ingress" yaml path is not a good choice as there are
services such neutron that use this path to define backend service,
named "server", options. We propose moving it to the root of the
path "network".
Change-Id: If98d6555a9c012872d3fb1a38b370a3195ea49ab
Based on spec in openstack-helm repo,
support-OCI-image-registry-with-authentication-turned-on.rst
Each Helm chart can configure an OCI image registry and
credentials to use. A Kubernetes secret is then created with these
info. Service Accounts then specify an imagePullSecret specifying
the Secret with creds for the registry. Then any pod using one
of these ServiceAccounts may pull images from an authenticated
container registry.
Change-Id: Iebda4c7a861aa13db921328776b20c14ba346269
The change makes sure that the script runs idempotently
and if run twice in a row no extra work is going to be done
Change-Id: Ifc055b32528bc4702b480f2de92c544578c96d73
The major reason for the addition of this feature is to facilitate
an upgrade to the Pacific Ceph release, which now requires the
require-osd-release flag to be set to the proper release in order
to avoid a cluster warning scenario. Any Ceph command can be run
against the cluster using this feature, however.
Change-Id: I194264c420cfda8453c139ca2b737e56c63ef269
This change allows creating a single ingress resource using the
public fqdn of the service, instead of two (cluster and namespace)
that is currently the case. Every openstack-helm chart can have a
network.server.ingress.use_external_ingress_controller boolean
field to choose the creation of a single ingress resource
(ingressName-namespace-fqdn).
Signed-off-by: Yanos Angelopoulos <yanos@admin.grnet.gr>
Change-Id: I46da850fccc3fee76595a2e6c49d51197a282c3e
Running 'tox -e releasenotes' fails with:
ERROR: You must give at least one requirement to install
Added missing doc/requirements.txt dep to tox.ini, similar
to recent openstack-helm patch.
Trivialfix
Change-Id: I01f82cdd92828ad0c99c343a1261f57562be3041
Python needs to be set to python2 before checking for python3 to prevent
certain test framework errors.
Change-Id: Ifd1ed35160338688d3c723c055ca75cd999e46e0
Readiness probe that we currently have does not help with restarting a
pod that got stuck in a transfer state reported by
wsrep_local_state_comment.
root@mariadb-server-2:/# mysql_status_query wsrep_ready
OFF
root@mariadb-server-2:/# mysql_status_query wsrep_connected
ON
root@mariadb-server-2:/# mysql_status_query wsrep_cluster_status
non-Primary
root@mariadb-server-2:/# mysql_status_query wsrep_local_state_comment
Transfer
So the idea is to add a liveness probe that will take care of this.
Change-Id: I2ccecc75349667fe19c6f7f9dccc2dbbd17d0a5e
