[RALLY] Running Rally tasks and Tempest tests in multi-user context
Hi openstack-discuss,
I'm currently using Rally v3.4.0 to test our OpenStack Zed platform.
Rally has been deployed on a dedicated virtual machine and Rally tasks and Tempest tests, launched on this machine by Rundeck, run pretty well.
Now, I wish every OpenStack team member could launch whatever scenario or test he wants, when he wants, for example after having applied a service configuration change on the staging platform. And a question is arising: can several users launch different Rally scenarios or Tempest tests at the same time, from their own Linux account/environment, using the same Rally, the one which is deployed on the dedicated machine ?
Thanks and best regards,
Jean-Francois
hi!
чт, 3 авг. 2023 г. в 18:22, Taltavull Jean-François < jean-francois.taltavull@elca.ch>:
Hi openstack-discuss,
I’m currently using Rally v3.4.0 to test our OpenStack Zed platform.
Rally has been deployed on a dedicated virtual machine and Rally tasks and Tempest tests, launched on this machine by Rundeck, run pretty well.
Now, I wish every OpenStack team member could launch whatever scenario or test he wants, when he wants, for example after having applied a service configuration change on the staging platform.
And a question is arising: can several users launch different Rally scenarios or Tempest tests at the same time, from their own Linux account/environment, using the same Rally, the one which is deployed on the dedicated machine ?
Rally Task framework itself does not have limitations for parallel executions. But it is worth considering a few nuances: - Database backend. Rally is configured to use SQLite by default. It does not support simultaneous write operations, which can be an issue for using a dedicated rally instance for running parallel tasks. Switching to MySQL/Postgres should not have such an issue. - File descriptions. If you run too many parallel tasks with a huge number of parallel iterations, you may face default linux limitation of open file descriptors.
I never ran Tempest simultaneously for the same cloud, so I cannot guarantee that there are no bugs, but it should be ok in general.
Thanks and best regards,
Jean-Francois
Thanks a lot for this clear answer Andrey !
All the best, JF
From: Andriy Kurilin andr.kurilin@gmail.com Sent: dimanche, 6 août 2023 20:38 To: Taltavull Jean-François jean-francois.taltavull@elca.ch Cc: openstack-discuss@lists.openstack.org Subject: Re: [RALLY] Running Rally tasks and Tempest tests in multi-user context
EXTERNAL MESSAGE - This email comes from outside ELCA companies.
hi!
чт, 3 авг. 2023 г. в 18:22, Taltavull Jean-François <jean-francois.taltavull@elca.chmailto:jean-francois.taltavull@elca.ch>: Hi openstack-discuss,
I’m currently using Rally v3.4.0 to test our OpenStack Zed platform.
Rally has been deployed on a dedicated virtual machine and Rally tasks and Tempest tests, launched on this machine by Rundeck, run pretty well.
Now, I wish every OpenStack team member could launch whatever scenario or test he wants, when he wants, for example after having applied a service configuration change on the staging platform. And a question is arising: can several users launch different Rally scenarios or Tempest tests at the same time, from their own Linux account/environment, using the same Rally, the one which is deployed on the dedicated machine ?
Rally Task framework itself does not have limitations for parallel executions. But it is worth considering a few nuances: - Database backend. Rally is configured to use SQLite by default. It does not support simultaneous write operations, which can be an issue for using a dedicated rally instance for running parallel tasks. Switching to MySQL/Postgres should not have such an issue. - File descriptions. If you run too many parallel tasks with a huge number of parallel iterations, you may face default linux limitation of open file descriptors.
I never ran Tempest simultaneously for the same cloud, so I cannot guarantee that there are no bugs, but it should be ok in general.
Thanks and best regards,
Jean-Francois
-- Best regards, Andrey Kurilin.
participants (2)
-
Andriy Kurilin
-
Taltavull Jean-François