This will configure the 40Gbit interfaces and FRR daemon with OpenFabric routing. Our CEPH setup uses a private IPv6 subnet for inner cluster communication.

fdcc:a182:4fed::/64

You can check Mesh network status by launching FRR shell vtysh and then typing show openfabric topology

root@pve91:~# vtysh
pve91# show openfabric topology
IS-IS paths to level-2 routers that speak IPv6
Vertex                  Type          Metric  Next-Hop  Interface   Parent
 ------------------------------------------------------------------------------
pve91                                                                         
fdcc:a182:4fed::91/128  IP6 internal  0                             pve91(4)  
pve93                   TE-IS         10      pve93     enp161s0    pve91(4)  
pve90                   TE-IS         10      pve90     enp161s0d1  pve91(4)  
fdcc:a182:4fed::93/128  IP6 internal  20      pve93     enp161s0    pve93(4)  
fdcc:a182:4fed::90/128  IP6 internal  20      pve90     enp161s0d1  pve90(4)

Setup CEPH packages on all nodes

pveceph install --repository no-subscription --version squid

CEPH init

pveceph init --network fdcc:a182:4fed::/64

Create CEPH monitors on each node
```
pveceph mon create
```
Also create CEPH managers on each node
```
pveceph mgr create
```

Create OSD daemons for each disk on all nodes

NVMe drives will get 2 OSD daemons per disk for better IOPS

pveceph osd create /dev/nvme0n1 --crush-device-class nvme --osds-per-device 2

HDD-s will get just 1

pveceph osd create /dev/sdX --crush-device-class hdd

Create CRUSH Maps

We want to separate out HDD and NVMe storage into different storage buckets.

Default replicated_rule would put datablock on all of the available disks
```
# ceph osd crush rule create-replicated <rule-name> <root> <failure-domain> <class>
ceph osd crush rule create-replicated replicated_nvme default host nvme
ceph osd crush rule create-replicated replicated_hdd default host hdd
```
NB: Using default replicated_rule for ANY CEPH Pool will result in Placement Group (PG) Autoscaler not working as it cant properly calculate how much space is available in CEPH due to different device classes we are using
Create CEPH Pools for VM disk images

This is done in individual node Ceph -> Pools configuration

NB: Under advanced, select correct Crush Rule (nvme or hdd)
Create CephFS Storage pool for ISO images

First create metadata server on each node
```
pveceph mds create
```
Then on one of the individual nodes create a CephFS.

After that is done you can modify under Pools change the cephfs_data and cephfs_metadata Crush rules to use NVMe drives.

CEPH NUMA Pinning

This helps a bit with read latency (482.28us vs 437.22us)

Inside hwloc-nox package there a programm called hwloc-ls that will visualize connected hardware and NUMA nodes. In our case Ceph network interface and NVMe drive are both connected to the same NUMA node. We can use hwloc-calc -I core os=nvme0n1 to get a list of CPU cores attached to the NVMe drive.

# hwloc-calc -I core os=nvme0n1
8,9,10,11,12,13,14,15

From that output we can create a systemd override file for ceph-osd@<ID> daemons.

systemctl edit ceph-osd@0

And then paste

[Service]
CPUAffinity=8,9,10,11,12,13,14,15
NUMAPolicy=default
NUMAMask=8,9,10,11,12,13,14,15

After restarting the OSD you should see in numastat ceph-osd that OSD is contained to mostly single node.

Here are bunch of example fio benchmark commands that can be used to verify this change

https://docs.oracle.com/en-us/iaas/Content/Block/References/samplefiocommandslinux.htm