> 
> Deployment process did create both _hdd_ & _ssd_ classes.
> I've read somewhere in docs that device classes are more of a concept rather 
> than a real thing

Hmm.  I've written that the *names* are arbitrary, so one can assign different 
ones, but it's best to only do that if you have a specific reason.

Say you have NVMe SSDs, SATA SSDs, and SATA HDDs, and enough of each that you 
want to constrain which pool goes on which.

By default, without anything exotic, the HDDs will be reported by the kernel as 
rotational, so OSDs on them will get the device class "hdd".
The SATA and NVMe SSDs will be reported by the kernel as nonrotational, so they 
will get the device class "ssd".  With both, one might, for example wish to 
place pools:

* CephFS .meta : NVMe SSDs
* CephFS first .data: SATA SSDs
* CephFS additional .data pools: HDDs

In that case one might assign the NVMe SSDs a CRUSH device class of "nvme" and 
define a CRUSH rule that only uses those.  In a cluster with mixed media, one 
almost never wants a given pool to use more than one, like you wouldn't want a 
CephFS .meta pool on both HDDs and SSDs, or the data pools eating the SSD 
capacity.


> - which would confirm since my cluster is a lab, all in virtual env, all 
> storage devices/drives all the same "type" and yet these two classes were 
> auto-created (and used by default if I remember, at osd-creation time)

How is your cluster deployed?  Something would have to go out of its way to 
create a device class called MONERO.  I suspect that you might have 
misinterpreted a configuration item.

Virtual storage devices as seen by Ceph may be classed differently than one 
would want, usually as non-rotational when there are SSDs behind the 
abstraction.  This does matter in some ways, as some configuration options have 
_hdd and _ssd variants that are used depending on the rotational attribute.  In 
such cases a udev rule can coerce what you want.

Are you using Proxmox or Rook or something to deploy?


> 
> _______________________________________________
> ceph-users mailing list -- [email protected]
> To unsubscribe send an email to [email protected]
_______________________________________________
ceph-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to