Description of problem: Older Prometheus stack install (and grafana) has been deprecated in favor of the openshift-monitoring stack. Since the older prometheus install was a Tech Preview, and is no longer necessary it should be removed.
PR: https://github.com/openshift/openshift-ansible/pull/10107/files
The change looks fine, but it needs a doc bug, mentioning that existing Prometheus/Grafana needs to be uninstalled first, as these won't be automatically removed during upgrade
PR for backport to 3.11.x: https://github.com/openshift/openshift-ansible/pull/10142
Change is merged to 3.11
The fix is OK, but the following openshift_prometheus parameters should also be removed # grep -r "openshift_prometheus" roles/lib_utils/action_plugins/sanity_checks.py: 'openshift_prometheus_alertbuffer_storage_kind', roles/lib_utils/action_plugins/sanity_checks.py: 'openshift_prometheus_alertmanager_storage_kind', roles/lib_utils/action_plugins/sanity_checks.py: 'openshift_prometheus_storage_kind') Binary file roles/lib_utils/action_plugins/sanity_checks.pyc matches roles/openshift_facts/defaults/main.yml:openshift_prometheus_storage_volume_name: 'prometheus' roles/openshift_facts/defaults/main.yml:openshift_prometheus_storage_volume_size: '10Gi' roles/openshift_facts/defaults/main.yml:openshift_prometheus_storage_access_modes: roles/openshift_facts/defaults/main.yml:openshift_prometheus_storage_create_pv: True roles/openshift_facts/defaults/main.yml:openshift_prometheus_storage_create_pvc: False roles/openshift_facts/defaults/main.yml:openshift_prometheus_alertmanager_storage_volume_name: 'prometheus-alertmanager' roles/openshift_facts/defaults/main.yml:openshift_prometheus_alertmanager_storage_volume_size: '10Gi' roles/openshift_facts/defaults/main.yml:openshift_prometheus_alertmanager_storage_access_modes: roles/openshift_facts/defaults/main.yml:openshift_prometheus_alertmanager_storage_create_pv: True roles/openshift_facts/defaults/main.yml:openshift_prometheus_alertmanager_storage_create_pvc: False roles/openshift_facts/defaults/main.yml:openshift_prometheus_alertbuffer_storage_volume_name: 'prometheus-alertbuffer' roles/openshift_facts/defaults/main.yml:openshift_prometheus_alertbuffer_storage_volume_size: '10Gi' roles/openshift_facts/defaults/main.yml:openshift_prometheus_alertbuffer_storage_access_modes: roles/openshift_facts/defaults/main.yml:openshift_prometheus_alertbuffer_storage_create_pv: True roles/openshift_facts/defaults/main.yml:openshift_prometheus_alertbuffer_storage_create_pvc: False roles/openshift_logging_elasticsearch/defaults/main.yml:openshift_logging_elasticsearch_prometheus_sa: "system:serviceaccount:{{openshift_prometheus_namespace | default('openshift-metrics')}}:prometheus" version: # rpm -qa | grep ansible openshift-ansible-playbooks-3.11.31-1.git.0.d4b5614.el7.noarch openshift-ansible-docs-3.11.31-1.git.0.d4b5614.el7.noarch openshift-ansible-roles-3.11.31-1.git.0.d4b5614.el7.noarch openshift-ansible-3.11.31-1.git.0.d4b5614.el7.noarch ansible-2.6.6-1.el7ae.noarch
I think we should not remove prometheus playbooks from 3.11 branch, prometheus will be officially deprecated since 4.0, if there is update on openshift-ansible 3.11, customers can not use the openshift-ansible 3.11 to deploy prometheus. PR should be only merged to 4.0 and above.
Variable cleanup PR for master: https://github.com/openshift/openshift-ansible/pull/10590 PR for 3.11: https://github.com/openshift/openshift-ansible/pull/10616
Paul Prometheus parameters are removed from openshift-ansible:v3.11 # rpm -qa | grep openshift-ansible openshift-ansible-playbooks-3.11.44-1.git.0.11d174e.el7.noarch openshift-ansible-roles-3.11.44-1.git.0.11d174e.el7.noarch openshift-ansible-3.11.44-1.git.0.11d174e.el7.noarch openshift-ansible-docs-3.11.44-1.git.0.11d174e.el7.noarch We have v3.11 older Prometheus images and tested them, and fixed some v3.11 bugs, for example: Bug 1569400 and Bug 1608288, maybe some users are using the older Prometheus now. I am strongly recommend to retain the older prometheus install from openshift-ansible v3.11 installer, and only remove them since 4.0 WDYT?
assign back, should not merge PR to v3.11 since customers are using the older prometheus
As far as I know, the old Prometheus tech preview stack has been removed shortly after the 3.11.0 release since it has been replaced by the cluster monitoring stack. Furthermore, the tech preview was never supported either and everyone who is using the TP should consider to move into the cluster monitoring stack. I don't understand exactly what your concerns are since it never came up with customers that we removed the playbook from that release. Can you elaborate?
(In reply to Christian Vogel from comment #12) > As far as I know, the old Prometheus tech preview stack has been removed > shortly after the 3.11.0 release since it has been replaced by the cluster > monitoring stack. Furthermore, the tech preview was never supported either > and everyone who is using the TP should consider to move into the cluster > monitoring stack. > > I don't understand exactly what your concerns are since it never came up > with customers that we removed the playbook from that release. > > Can you elaborate? The first release of old Prometheus is v3.11.16, now the latest release will be 3.11.43, and we have all the old prometheus images released to customer, including prometheus-alert-buffer image. I thought we should support it for consistency's sake。 Since the old Prometheus is only tech preview and we will stop support the old Prometheus, we will remove the prometheus-alert-buffer image from our Errata Will set this bug to VERIFIED and close it.
Per Comment 9 and Comment 13, set to VERIFIED
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHBA-2019:0024