Graphing issues with roundrobin load balancing prometheus in grafana

[Will Szumski]

When grafana queries Prometheus, it will be load balanced across all of the Prometheus server instances. Because these instances are are scraping independently the timestamps for the metrics can be slightly off. This causes data points to appear slightly misaligned from the previous query.

Possible solutions:

- Have one primary server and the rest configured as backups
- Sticky sessions

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.opendev.org/c/openstack/kolla-ansible/+/790936

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.opendev.org/c/openstack/kolla-ansible/+/790936
Committed: https://opendev.org/openstack/kolla-ansible/commit/6906b275ef58c5d35bef73fd64054925e9a7fa15
Submitter: "Zuul (22348)"
Branch: master

commit 6906b275ef58c5d35bef73fd64054925e9a7fa15
Author: Will Szumski <email address hidden>
Date: Mon Dec 14 18:29:21 2020 +0000

    Switch prometheus to active/passive mode

    This uses the same approach as the mariadb role (and others).

    Closes-Bug: #1928193
    Co-Authored-By: John Garbutt <email address hidden>
    Change-Id: I79a7a8c80327cfd9ef31d17fe71f450a181a638c

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/kolla-ansible 15.0.0.0rc1 release candidate.