Hey I was just messing around with network settings on a proxmox+ceph cluster and I managed to trigger reboots in both nodes that I was messing with.
The changes I had made caused the nodes to not be able to see each other properly (I basically moved the corosync onto a balance-alb bond which in hindsight was not a good idea to put it mildly) and seem to have caused a hard crash of some sort.
I'm trying to look at journalctl to see if there is any useful information there, is this useful to anyone or should I just move on?
Some background -
Cluster is a former vxrails cluster (3x dell S570) with 4 10GbE nics per node, the switch they're connected to freezes every time the network admin tries to create an LACP bond and as it is a core switch she is (justifiably) not happy to do a lot of playing around.
Situation before changes:
bond0 - nic2-4, balance-alb
ceph0 - bridge using bond0 with SDN Q-in-Q and SDN vlan on it.
vmbr0 - bridge using nic1
vmbr0.tag - management of proxmox clutser
(ceph lives on one of the Q-in-Q vlans)
After changed [del vmbr0.tag, del vmbr0, add ceph0.tag]:
bond0 - nic1-4, balance-alb
ceph0 - bridge using bond0 with SDN Q-in-Q and SDN vlan on it.
ceph0.tag - management of proxmox clutser
Change to stabilize everything again:
bond0 - nic1-4, balance-alb
ceph0 - bridge using bond0 with SDN Q-in-Q and SDN vlan on it.
nic1.tag - management of proxmox clutser
The changes I had made caused the nodes to not be able to see each other properly (I basically moved the corosync onto a balance-alb bond which in hindsight was not a good idea to put it mildly) and seem to have caused a hard crash of some sort.
I'm trying to look at journalctl to see if there is any useful information there, is this useful to anyone or should I just move on?
Some background -
Cluster is a former vxrails cluster (3x dell S570) with 4 10GbE nics per node, the switch they're connected to freezes every time the network admin tries to create an LACP bond and as it is a core switch she is (justifiably) not happy to do a lot of playing around.
Situation before changes:
bond0 - nic2-4, balance-alb
ceph0 - bridge using bond0 with SDN Q-in-Q and SDN vlan on it.
vmbr0 - bridge using nic1
vmbr0.tag - management of proxmox clutser
(ceph lives on one of the Q-in-Q vlans)
After changed [del vmbr0.tag, del vmbr0, add ceph0.tag]:
bond0 - nic1-4, balance-alb
ceph0 - bridge using bond0 with SDN Q-in-Q and SDN vlan on it.
ceph0.tag - management of proxmox clutser
Change to stabilize everything again:
bond0 - nic1-4, balance-alb
ceph0 - bridge using bond0 with SDN Q-in-Q and SDN vlan on it.
nic1.tag - management of proxmox clutser