Better utilization of a 3node ceph cluster

[FSNaval]

Hello everyone,

I currently have a 3node cluster running ceph with two data pools. One data pool with NVMEs for VMs; One data pool with HDD for bulk data.

I have deployed a few VMs on the cluster and have been running smoothly and stable for the last two years without a hiccup.

The nodes are not similar in their specs, namely I have an i59400 with 48GB RAM, one i512400 with 64GB RAM and one i313100 with 32GB of RAM.

One of the VM sits on the i512400 and runs my NAS as well as a good amount of docker services.

I am thinking of how to better utilize my current hardware and I am thinking of trying docker swarm, since the most beefier machine takes almost all the load and the other ones are almost running idle unless something happens to the bog machine and high availability kicks in.

PS: The other machines are able to handle the load of the big one but this will lead them to hit 95% RAM Usage which is not ideal.

The questions I have is How will I configure my shared storage? I am thinking of cephfs but:

I have not touched it in the past and for accessing the data, I use windows and macOS and I don’t know how to access cephFS from them. I saw some YouTube videos for windows but nothing for Mac.

Are there any other alternatives I can look into that will help me utilize my hardware better?

I can always leave things as is , since they are working flawlessly for the last two years.

Thank you all for your replies

 

[ness1602]

Cephfs is not meant for nas, like small files,so don't do that.

 

[FSNaval]

Cephfs is not meant for nas, like small files,so don't do that.

Thank you for your reply.

Another thought that crossed my mind is that sonce i already have a dedicated pool for my VMs on ceph, i could:

• Create three new VMs to be used as my docker swarm cluster (each vm sitting on its own machine)
• During VMs creation, to overprovision the hard disk of each VM (for example, 512GB of disk for OS)
• Create a specific folder at root directory to be used for docker volumes, for example /dockervolumes

Since all VMs will be sitting on distributed storage controlled by ceph itself, if any machine goes down, ceph will migrate this machine to another node by itself and thus my swarm cluster will remain operational.

Do you think the above will work?