Question on Ceph and CephFS - Advice.

[Seed]

I am currently using ceph and have a few pools that use RBD storage type. One of these pools is rather large and responsible for holding onto media only, or will be. This has 10s of thousands of pictures, videos, music.

Should I create a cephFS fs on the large media disks? I'm not sure how or what to do here. I am making a large disk volume on these large disks, that will attach to Virtual Machines that run on SSDs. Those reside on the RBD disks.

Ideally I'd like to pass it through to the VM i think? Not sure.

Is that right?

 

[Alwin]

I am sure, you want to share these files over the network. Possibly with different clients (Lin/Mac/Win). Best use a NFS/SMB server (as VM) and host it over that. It is of course possible to create a CephFS on the existing Ceph cluster and connect the NFS/SMB server to it. But this is way more complex.

 

[Seed]

Best use a NFS/SMB server (as VM) and host it over that.

So are you saying just create a vm disk and mount it in the VM? My current approach is to do that and then rsync the data onto it from a NAS. Feels kludgey but maybe it's the right way? Then use docker to access the volume from the VM.

 

[Alwin]

So are you saying just create a vm disk and mount it in the VM? My current approach is to do that and then rsync the data onto it from a NAS. Feels kludgey but maybe it's the right way? Then use docker to access the volume from the VM.

I don't understand, why/where you want to use docker, but yes.

 

[Seed]

I don't understand, why/where you want to use docker, but yes.

I have docker and docker swarm in VMs, and portainer already on omv and looking to spread this out across shared storage backed VMs with docker inside. This way I can move VMs with the containers freely and not lose anything and eventually automate the VMs being created/added I think. It's a lot easier to throw away containers than VMs. With docker we can iterate changes extremely fast. A lot easier to throw away VMs over bare metal, so just looking at getting more experience managing what's likely pretty typical infrastructure in larger enterprises from top to bottom.