VM failed and won't start. Displaying "An emulator required to run this VM failed to start"

rustylh

xcp-ng iscsiadm -m session -P 3 Attached scsi disk sdb State: offline

olivierlambert

There's something fishy about the connectivity to your storage.

rustylh

@olivierlambert I am looking into that. I just do not have much experience with the networking.

rustylh

I found that the SR has two NICs bonded to two different hardware switches on the iSCSI LUN. Then each host has a management network(Bonded NICS on hosts) and VM network (Bonded NICS). Should I break the SR bond and just go with one NIC? If so, I assume I would be best to shutdown the VMs until I have one NIC on SR reconfigured?

rustylh

Additional finding. The two switches (LAG) are Cisco with STP enabled.

rustylh

I am still getting " sd 7:0:0:1: rejecting I/O to offline device" taking all VMs on host down. I then shutdown the VM and power them on, on another Host. I have 4 host in a pool connected iSCSI to a Synology LUN. What is the best way to troubleshoot network issues on xcp-ng.
Thank you

rustylh

Just adding another note. When I reboot the host everything is fine then the issues pops back up. There doesn't seem to be any consistency as far as timeframe in between. It just happens randomly

rustylh

Any help is greatly appreciated.

olivierlambert

That's not a trivial problem that you have, it seems to be intermittent and could be anything between host and the storage. I assume you can see the issue as soon you have the rejecting I/O to offline device right?

rustylh

Yes

olivierlambert

At least we know there's a root cause: the device becomes offline. Could it be a power setting or something on the device itself?

rustylh

Well, it doesn't happen to every host at the same time, it's just random.

rustylh

What log is best to look at for network disconnects?

olivierlambert

Well dmesg is fine, I would also check the switch logs if you have anything happening on the ports where you host are connected. Since you can find when it's happening with dmesg, I would use that to correlate with anything else happening few minutes before.

rustylh

dmesg -T is only showing these from 00:37 on.

[Mon Jun 30 05:55:10 2025] print_req_error: I/O error, dev sdb, sector 0
[Mon Jun 30 05:55:10 2025] sd 7:0:0:1: rejecting I/O to offline device

olivierlambert

Then try to find anything happening around that time on other hosts, equipment, storage and so on.