no HA failover with loss of iSCSI

P

philRAL

New Member
Aug 12, 2024
6
1
3
Hi,

we have been running some HA testing on a new 5 node cluster. Each node has dual redundant connections to a data network (bonded) , plus dual redundant connections to an iSCSI network (multipathed).

We did some cable pull tests today to check HA bevaviour. Networking was as expected - loss of 1 cable was handled by bonding and loss of both cables caused an HA failover for the VMs on the hypervisor.

When we pulled the iSCSI cables, 1 cable was handled by the multipathing as expected. However, when we pulled the other cable (so the VM had lost its system disk effectively) there was no failover.

Is this expected behavior for Proxmox, or is there some configuration setting we can make to failover on loss of the iSCSI storage?

Thanks
P
 
ok thanks for clarifying that. I guess we could use some manual monitoring on the host to initiate an ha-manager command to put the node into maintenance state.
 
philRAL said:
ok thanks for clarifying that. I guess we could use some manual monitoring on the host to initiate an ha-manager command to put the node into maintenance state.
Click to expand...
Is this a real use case? I've never seen this in 15y+ on Linux with multipath, that only the storage on one node fails besides a whole node failure. Storage and underlying network (ethernet or FC) can and will fail, but the node itself? Never experienced this.
 
  • Like
Reactions: Johannes S
You must log in or register to reply here.
Top Bottom