| mhen | tafkamax: that is one way to do it; you can have it back up local paths from within the VM | 09:57 |
|---|---|---|
| mhen | however there is also a mode where it uses OpenStack credentials of the tenant (project) and talks to the OpenStack APIs (Nova, Cinder, Glance) to backup volumes and VMs through snapshots and images, i.e., outside of the VM through the infrastructure | 09:58 |
| mhen | this does currently require you to run scheduler + agent on a per-project basis from somewhere that has API access | 09:59 |
| tafkamax | So in essence you need a controller VM, but that is not per say on the ctrl plane. Rather it is per tenant. | 10:11 |
| tafkamax | Well we are using rsnapshot at the moment and that is why we are looking at freezer. | 10:11 |
| tafkamax | I don't know the whole architecture of freezer. I presume if an orchestrator would to use freezer something needs to be set up (kolla as the orchestrator). But the scheduler is up for the operator themselves to setup. | 10:12 |
| tafkamax | Or the end users. | 10:12 |
| noonedeadpunk | as of today - yes | 10:49 |
| tafkamax | Ok, so this implies that you want to change this behavior? | 10:50 |
| tafkamax | if you say as of today? | 10:50 |
| mhen | tafkamax: a new service mode using a centralized scheduler is in the works by noonedeadpunk et al | 17:37 |
| mhen | see https://specs.openstack.org/openstack/freezer-specs/specs/2025.1/approved/centralized-scheduler.html | 17:38 |
| *** mhen_ is now known as mhen | 23:40 | |
Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!