Laurent, thank you very much for all this help. My current Openstack platform is built for labs, so not saving data is not a problem. We show a lot of interesting things on this cluster, the students really appreciate it. They also appreciate labs (virtual or on physical servers) where we build a small openstack (first all-in-one, with keystone and AD link, with tls, then multinode, designate, etc.). But making the right hardware choices for a new cluster... it's not the same thing at all. It's really difficult. If we don't make the right choices, it's for a long time. So explained in your message, I will turn to ephemeral storage on SSD disks on the cluster servers. I know the principles of raw and qcow2 images, I also have a cache set up for using these images. But I'm using it wrong, I know. This will be corrected on the next cluster. Franck
Le 13 mars 2024 à 00:25, Laurent Dumont <laurentfdumont@gmail.com> a écrit :
Bonjour Franck,
Good to know! If the end goal is a lab for folks to get familiar with Openstack, it makes sense to keep the platform.
From an image creation perspective, you might be overloading the IO on the SAN. Like Peter mentioned, .raw and .qcow2 images behave differently when it comes to creation/clone. Are you creating an instance from a snapshot? Are you creating an instance from an image? What is the format used From a speed perspective : If storage reliability/redundancy is not a concern. If losing data is not the end of the world. If you have local storage on the compute and it's based on SSD/NVME You could try ephemeral storage? It's basically placing the qemu instance disk directly on the compute. No need for ceph/iscsi or more servers, just add a couple of drives to the compute itself. There is some level of instance caching --> https://docs.openstack.org/nova/latest/user/support-matrix.html#operation_ca... Laurent
On Tue, Mar 12, 2024 at 4:13 PM Franck VEDEL <franck.vedel@univ-grenoble-alpes.fr <mailto:franck.vedel@univ-grenoble-alpes.fr>> wrote:
Bonjour Laurent, and thanks a lot for your help.
Are you running out of vCPU/RAM due to too many VMs?
Yes… it’s one of my problems.
Have you looked at other virtualization platforms like Proxmox?
I know Proxmox well, I also use it for other labs. But here, it's about creating networks, routers, connecting them, taking control remotely, playing with security groups, using Designate.
I'm already going to stop using volumes so much, which are not useful to me most of the time. What I'm looking for, in addition to the ability to manage instances for my 200 students, is for it to be fast, for example I'm looking to optimize the time it takes to create Windows instances. When there are around thirty of them at the same time, that poses problems for me.
I'm still studying all of this. Thank you anyway.
Franck VEDEL Dép. Réseaux Informatiques & Télécoms IUT1 - Univ GRENOBLE Alpes 0476824462 Stages, Alternance, Emploi.
Le 12 mars 2024 à 04:08, Laurent Dumont <laurentfdumont@gmail.com <mailto:laurentfdumont@gmail.com>> a écrit :
Bonjour Franck!
It would be interesting to narrow down where you are reaching the limits of Openstack : Are you running out of vCPU/RAM due to too many VMs? Are you seeing CPU steal time due to oversubscribing the hosts? Are the VMs oversized? Are you running out of capacity on the network side? Are you running out of capacity on the storage side? Is the Openstack control plane reaching its limits? That would be surprising considering the amount of computes you have. At such a "small" scale, some of the Openstack overhead might be tricky to deal with. Running Controllers + Compute + Storage on the same server typically means a "hyperconverged" setup. I am not sure if there are community run deployments that will support it out of the box. It's typically complex to deploy/support/maintain and part of a commercial support package. Have you looked at other virtualization platforms like Proxmox?
On Thu, Mar 7, 2024 at 9:36 AM Franck VEDEL <franck.vedel@univ-grenoble-alpes.fr <mailto:franck.vedel@univ-grenoble-alpes.fr>> wrote:
Good morning, I currently have an Openstack cluster made up of 3 nodes, an iscsi bay (10T), 576 G of Ram, 10T, 288vcpus. This cluster is used by around 150 students, but is reaching its limits. Having obtained a budget to set up a larger cluster, I am wondering about the choice of the number of nodes, their role (how many controllers, network, compute, etc.) and above all what solution for storage. Let's imagine a budget to buy 6 servers with good capacities, is the right choice Ceph storage (with cinder and rdb?) on the Openstack cluster nodes? Do we need 3 servers for a Ceph cluster and 3 for the Openstack part (in this case I lose capacity for the "compute" part)... I don't know what the right choices are and above all, I have a little afraid of going in the wrong directions. Could any of you guide me, or give me links to sites that could help me (and that I haven't seen). Thanks in advance
Franck VEDEL Dép. Réseaux Informatiques & Télécoms IUT1 - Univ GRENOBLE Alpes 0476824462 Stages, Alternance, Emploi.