Description of problem: Gnocchi use s a lot of IOPs and thus can bog down any other ceph pools co-located on the same ceph storage that tripleo/director deploys. We should be able to determine which ceph storage nodes are going to store which pool. Ideally we can put Gnocchi metrics pool on its own set of storage nodes so the IOPs into the metrics pool doesn't slow down the other pools such as vms, images, backups, volumes, etc. Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: 1. 2. 3. Actual results: Expected results: Additional info: I can include some graphs/data surrounding % io util on disks at specific scales with a specific Gnocchi archive-policy if that is needed.
You can specify a pool for each project, but IIUC you want to have pools being created for different Ceph storage, which is not the case anymore. So that's has nothing to do with telemetry, since we just need a pool name. The rest is Ceph deployment/configuration options.
I read this as a request to have director configure the pool that Gnocci consumes up to Ceph specs. Seb,Josh can you indicate what is the next step, if any?
(In reply to Federico Lucifredi from comment #2) > I read this as a request to have director configure the pool that Gnocci > consumes up to Ceph specs. That sounds like the next step to me.
Correct, OSPd needs to add a pool for "metrics" and apply the appropriate configuration for Gnocchi.
In OSP13/16 you can use director to deploy different tiers of storage for pools as described here: https://access.redhat.com/documentation/en-us/red_hat_openstack_platform/13/html/deploying_an_overcloud_with_containerized_red_hat_ceph/assembly_ceph-second-tier-storage As per bz 1816989, in newer versions of ceph-ansible available in RHCSv3/4 which map respectively to OSPv13/16 this will be simpler to deploy. So in the case of this bug your method would be to use CephPools to override gnocchi pool and apply the rule that matches with the faster devices (ssd) In bz 1793525 we'll be testing methods like this so I'm closing this bug as a duplicate of it. *** This bug has been marked as a duplicate of bug 1793525 ***