
FYI: I recently found that ceph performs quite well in vmware, when using the VM as a "direct" ceph client. (Our tests involved directly mounting an RBD device from the VM) So as long as you can treat the OS level as throwaway, and use ceph-to-the-vm for your actually important data, this might be also a good thing to use in ovirt. ----- Original Message ----- From: "Eyal Shenitzky" <eshenitz@redhat.com> To: "Sandro Bonazzola" <sbonazzo@redhat.com>, "Benny Zlotnik" <bzlotnik@redhat.com>, matthew@peregrineit.net Cc: "users" <users@ovirt.org> Sent: Wednesday, April 21, 2021 12:30:39 AM Subject: [ovirt-users] Re: n00b Requesting Some Advice Hi Matthew, Currently, in order to use Ceph in oVirt you have 2 options: 1. Ceph using ISCSI gateway - regular ISCSI storage domain with Ceph as the storage backend, supports all the regular operations [1]. 2. Using the new Managed Block Storage (Cinderlib integration) technical-preview - Create a Managed Block Storage domain that doesn't support all the operations that we have for the "regular" storage domain you can find more info here [2] and [3]. Each option has its benefits. [1] - [ https://access.redhat.com/documentation/en-us/red_hat_ceph_storage/3/html/bl... | https://access.redhat.com/documentation/en-us/red_hat_ceph_storage/3/html/bl... ] [2] - [ https://access.redhat.com/documentation/en-us/red_hat_virtualization/4.4/htm... | https://access.redhat.com/documentation/en-us/red_hat_virtualization/4.4/htm... ] [3] - [ https://www.ovirt.org/develop/release-management/features/storage/cinderlib-... | https://www.ovirt.org/develop/release-management/features/storage/cinderlib-... ] On Wed, 21 Apr 2021 at 10:12, Sandro Bonazzola < [ mailto:sbonazzo@redhat.com | sbonazzo@redhat.com ] > wrote: [ mailto:eshenitz@redhat.com | +Eyal Shenitzky ] any suggestion? Il giorno lun 12 apr 2021 alle ore 11:15 < [ mailto:matthew@peregrineit.net | matthew@peregrineit.net ] > ha scritto: Hi All, I need some "best practice" advice. We have a Ceph Storage Cluster (Octopus moving to Pacific) which we'd like to use with our new oVirt Cluster (all on CentOS 8 boxes). What I'd like to know is what is the "best" (ie recommended / best practice) way of doing this - via iSCSI, CFS, 'raw' RBD blocks, some other way I haven't read about yet, etc? I realise 'best' is a subjective term, but what I tend to do is do 'manual' installs so that I both actually understand what is happening (ie how things fit together - I pull apart and rebuild mechanical clocks and watches for the same reason) and also so I can '"Puppet-ise" the results for future use. This means that I am *not* necessarily looking for "quick and dirty" or "quick and easy" (ie, I have no trouble using the CLI and 'vim-ing' conf files as required) but I do want a solid, "best-practice" system when I'm done. So, can some please help? And also, would you mind pointing me towards the relevant documentation for the answer(s) supplied (yes, I *always* RTFM :-) ). Thanks in advance Dulux-Oz