[neutron][qa] Databases used on CI jobs
Hi, Recently in Neutron we merged patch [1] which caused error in Kolla-ansible [2]. It seems that issue happens on mariadb 10.1 which is used in broken Kolla-ansible job so I started checking what database is used on Neutron’s jobs as we didn’t found any issue with this patch. And it seems that in all scenario/api jobs we have installed mysql 5.7 - see [3] for example. And the question is - should we maybe switch somehow to Mariadb instead of Mysql? Should it be done globally or maybe it’s only “issue” in Neutron’s CI jobs? Or maybe we should have jobs with both Mariadb and Mysql databases? [1] https://review.opendev.org/#/c/677221 [2] https://bugs.launchpad.net/kolla-ansible/+bug/1841907 [3] https://2f275f27875ae066d2b6-00a9605a715176e6bd23401c23e881d7.ssl.cf2.rackcd... — Slawek Kaplonski Senior software engineer Red Hat
On 2019-08-30 10:07:14 +0200 (+0200), Slawek Kaplonski wrote: [...]
The unfortunate reality is that there are multiple relational database servers available for folks to try to use, and we can't reasonably run automated tests against them all. The TC's 2017-06-13 resolution to "Document current level of community database support" is the closest thing we have to guidance around this in recent history: https://governance.openstack.org/tc/resolutions/20170613-postgresql-status.h... At the time the challenge was that we couldn't reliably test everything against both MySQL and PostgreSQL, and the user survey showed an overwhelming majority of known deployments used MySQL. We used the terminology "MySQL family of databases" to include MariaDB and Galera (particularly because upstream support for MySQL had waned in the wake of the Oracle acquisition and MariaDB was a recent enough fork that it was basically a drop-in bug-compatible replacement which some distros had even taken to misleadingly relabeling as "MySQL"). That was more than two years ago. Fast-forward to the present and it seems MariaDB and MySQL have diverged noticeably, as this recent article suggests: https://blog.panoply.io/a-comparative-vmariadb-vs-mysql As such, it's probably time to revisit our database testing requirements. I expect some of the challenge here is that CentOS 7 (and presumably other RH-derived distros) have "replaced" MySQL with MariaDB. Since our Project Testing Interface requires that we test against Latest Ubuntu LTS, Latest CentOS Major, and Latest openSUSE Leap, that may be the best way to frame this discussion: https://governance.openstack.org/tc/reference/project-testing-interface.html... -- Jeremy Stanley
---- On Fri, 30 Aug 2019 15:07:14 +0700 Slawek Kaplonski <skaplons@redhat.com> wrote ----
I am not sure how many projects are effected in this. But I will suggest to go for later option. You can add a new neutron job with Mariadb and if you need you can add the same in devstack gate as experimental to test the things on demand. -gmann
Hi,
Thx. I will then add scenario periodic job with MariaDB. In same way like we have it done now for PostgreSQL job. I don’t want to add yet another scenario job to our check/gate queue which already have a lot of various jobs. Will it work for You?
— Slawek Kaplonski Senior software engineer Red Hat
participants (4)
-
Ghanshyam Mann
-
Jeremy Stanley
-
Mark Goddard
-
Slawek Kaplonski