Hello, In the Installation document, under "LVM simple" we illustrate the following for this second scenario: https://access.redhat.com/documentation/en-us/red_hat_ceph_storage/3/html-single/installation_guide_for_red_hat_enterprise_linux/#installing-a-red-hat-ceph-storage-cluster osd_objectstore: bluestore osd_scenario: lvm devices: - /dev/sda - /dev/sdb - /dev/nvme0n1 In the second scenario, when there is a mix of traditional hard drives and SSDs, the data is placed on the traditional hard drives (sda, sdb) and the BlueStore database (block.db) is created as large as possible on the SSD (nvme0n1). However, recently it was identified that you can use this same concept to apply to separating HDDs and making use of more then one nvme. Having said, this I am proposing something along the lines of the following as an idea: ================================================================================ "osd_objectstore: bluestore osd_scenario: lvm devices: - /dev/sda - /dev/sdb - /dev/nvme0n1 - /dev/sdc - /dev/sdd - /dev/nvme1n1 In the second scenario, when there is a mix of traditional hard drives and SSDs, the data is placed on the traditional hard drives (sda, sdb) and the BlueStore database (block.db) is created as large as possible on the SSD (nvme0n1). Additionally, traditional hard drives (sdc,sdd) will receive the data while nvme1n1 will respectively receive the block.db for these two respective disks." - Please let me know if you have any additional questions on this. Thanks!
When this docbug is implemented there should be a warning which captures the following idea: Be careful when using the proposed syntax to not run into bug 1732915. You can do the following and get the desired results: devices: - slow - slow - slow - slow - fast - slow - slow - slow - slow - fast bug ONLY if the slow devices are HDDs and the fast devices are SSDs. If the slow devices are SSDs and the fast devices are NVMe SSDs, then the above won't work because all of the devices will look like SSDs and then you'll get 10 OSDs where two of them are faster instead of 8 OSDs with a fast bluestore DB device. If you want to deploy in that scenario, then you need to directly create the LVMs in advance as described in https://access.redhat.com/solutions/3954161
(In reply to John Fulton from comment #1) > When this docbug is implemented there should be a warning which captures the > following idea: > > Be careful when using the proposed syntax to not run into bug 1732915. You > can do the following and get the desired results: > > devices: > - slow > - slow > - slow > - slow > - fast > - slow > - slow > - slow > - slow > - fast > > bug ONLY if the slow devices are HDDs and the fast devices are SSDs. I meant to say _but_ only if....
*** Bug 1732915 has been marked as a duplicate of this bug. ***
Updating the QA Contact to a Hemant. Hemant will be rerouting them to the appropriate QE Associate. Regards, Giri