Hey.
We have a 4 nodes ceph cluster and a host that runs the virtual machines. Each server has for the ceph network 2 gigabit cards. We would like to configure the Ceph network using LACP on an Unifi USW-24 switch. I did the configuration on the switch, 2 ports LACP Aggregating, Auto Negociate link speed auto and within the selected native vlan. On the Proxmox Node, I create a bond that has the IP configuration, bond mode 802.3ad, hash policy Layer 2+3.
On the same vlan was included an 1gbps port on the unifi switch that uplinks to the current ceph switch.
Now the weird part. I connected the server that runs the virtual machines on the unifi switch, in LACP mode. All goes, well, it can access the ceph cluster, virtual machines are up and runing. But when I move a ceph node in the same LACP network configuration, the ceph cluster is unable to conect to this one.
I just don't understand why the host that runs the vms is woking and the ceph hosts does not.
If I ping from the ceph node inside the LACP network, to the rest of the nodes, I have no packet loss or other conectivity problems.
Any ideea what would be wrong?
PS: Setup configuration is Proxmox 7.4.18, Ceph 17.2.7
Thanks.
We have a 4 nodes ceph cluster and a host that runs the virtual machines. Each server has for the ceph network 2 gigabit cards. We would like to configure the Ceph network using LACP on an Unifi USW-24 switch. I did the configuration on the switch, 2 ports LACP Aggregating, Auto Negociate link speed auto and within the selected native vlan. On the Proxmox Node, I create a bond that has the IP configuration, bond mode 802.3ad, hash policy Layer 2+3.
On the same vlan was included an 1gbps port on the unifi switch that uplinks to the current ceph switch.
Now the weird part. I connected the server that runs the virtual machines on the unifi switch, in LACP mode. All goes, well, it can access the ceph cluster, virtual machines are up and runing. But when I move a ceph node in the same LACP network configuration, the ceph cluster is unable to conect to this one.
I just don't understand why the host that runs the vms is woking and the ceph hosts does not.
If I ping from the ceph node inside the LACP network, to the rest of the nodes, I have no packet loss or other conectivity problems.
Any ideea what would be wrong?
PS: Setup configuration is Proxmox 7.4.18, Ceph 17.2.7
Thanks.
Last edited: