nvme connect-all: --nqn parameter

[BrianMer]

Hi,

I'm trying to connect to a subsystem NQN via nvme connect-all but the --nqn parameter is not working, event though it is taken into account by the module, leading to not being connected to the target:

Failed to write to /dev/nvme-fabrics: Input/output error
failed to add controller, error failed to write to nvme-fabrics device

I tried with /etc/nvme/discovery.conf and via CLI but this is the same result.

The nvme connect command, however, is working with the same parameter.

Is it a normal behavior of nvme connect-all even if the parameter is listed in the command usage?

Thanks!

[Mr-Bossman]

@BrianMer what version?

[BrianMer]

@Mr-Bossman nvme version 2.13 ; libnvme version 1.13 from the Debian 13 official repo

[igaw]

Could you post the output of nvme connect-all -vv --nqn ...? It should show the command string written into /dev/nvme-fabrics.

[BrianMer]

scan controller nvme0
warning: use hostid '44a35d9d-XXXX-XXXX-XXXX-XXXXXXXXXXXX' which does not match uuid in hostnqn 'nqn.2014-08.org.nvmexpress:uuid:f69d0000-XXXX-XXXX-XXXX-XXXXXXXXXXXX'
lookup subsystem /sys/class/nvme-subsystem/nvme-subsys0/nvme0
skipping namespace scan for ctrl nvme0
skipping path scan for ctrl nvme0
scan controller nvme1
lookup subsystem /sys/class/nvme-subsystem/nvme-subsys0/nvme1
failed to lookup subsystem for controller nvme1
failed to scan ctrl nvme1: No such device or address
scan subsystem nvme-subsys0
skipping namespace scan for subsys nqn.1994-11.com.samsung:nvme:PM9A1a:M.2:XXXXXXXXXXXXXX
warning: use hostid '44a35d9d-XXXX-XXXX-XXXX-XXXXXXXXXXXX' which does not match uuid in hostnqn 'nqn.2014-08.org.nvmexpress:uuid:f69d0000-XXXX-XXXX-XXXX-XXXXXXXXXXXX'
kernel supports: instance cntlid transport traddr trsvcid nqn queue_size nr_io_queues reconnect_delay ctrl_loss_tmo keep_alive_tmo hostnqn host_traddr host_iface hostid duplicate_connect disable_sqflow hdr_digest data_digest nr_write_queues nr_poll_queues tos keyring tls_key fast_io_fail_tmo discovery dhchap_secret dhchap_ctrl_secret tls concat recovery_delay 
connect ctrl, 'nqn=nqn.2000-02.com.mikrotik:xy_zab_raid6_array,transport=tcp,traddr=XX.XXX.X.X,host_traddr=XX.XXX.X.X,host_iface=XXX,trsvcid=4420,hostnqn=nqn.2014-08.org.nvmexpress:uuid:f69d0000-XXXX-XXXX-XXXX-XXXXXXXXXXXX,hostid=44a35d9d-XXXX-XXXX-XXXX-XXXXXXXXXXXX,discovery,ctrl_loss_tmo=600'
Failed to write to /dev/nvme-fabrics: Input/output error
failed to add controller, error failed to write to nvme-fabrics device

[igaw]

I think I see the problem. The hostnqn and hostid needs to have to use the same uuid. There is some logic code which does try to set this up correctly but I am not sure if libnvme 1.13 behaves correctly. This should be fixed with master though.

Anyway, a workaround is to provide the hostid via /etc/nvme/hostid or via the command line. Can you try this?

[BrianMer]

Yeah, I did not set up the hostid and the hostnqn, it has been done during installation of the nvme packet.
I tried via the parameter in CLI and the file /etc/nvme/hostid in order to provide the same UUID, but it still does not connect via nvme connect-all.

[igaw]

And what is the output of connect-all -vv? What are your command lines?

[BrianMer]

Here is the command line and the output both with and without the CLI parameter:
With the CLI parameter:

root@host:~# /usr/sbin/nvme connect-all --transport=tcp --traddr=XX.XXX.X.X --trsvcid=442
0 --host-iface=XXX --host-traddr=XX.XXX.X.X -n nqn.2000-02.com.mikrotik:xy_zab_raid6_array --host
id=f69d0000-xxxx-xxxx-xxxx-xxxxxxxxxxxx -vv
scan controller nvme0
warning: use hostid '44a35d9d-xxxx-xxxx-xxxx-xxxxxxxxxxxx' which does not match uuid in hostnqn 'nqn.2014-08.org.nvmexpress:uuid:f69d0000-xxxx-xxxx-xxxx-xxxxxxxxxxxx'
lookup subsystem /sys/class/nvme-subsystem/nvme-subsys0/nvme0
skipping namespace scan for ctrl nvme0
skipping path scan for ctrl nvme0
scan subsystem nvme-subsys0
skipping namespace scan for subsys nqn.1994-11.com.samsung:nvme:PM9A1a:M.2:XXXXXXXXXXXXXX
kernel supports: instance cntlid transport traddr trsvcid nqn queue_size nr_io_queues reconnect_delay ctrl_loss_tmo keep_alive_tmo hostnqn host_traddr host_iface hostid duplicate_connect disable_sqflow hdr_digest data_digest nr_write_queues nr_poll_queues tos keyring tls_key fast_io_fail_tmo discovery dhchap_secret dhchap_ctrl_secret tls concat recovery_delay 
connect ctrl, 'nqn=nqn.2000-02.com.mikrotik:xy_zab_raid6_array,transport=tcp,traddr=XX.XXX.X.X,host_traddr=XX.XXX.X.X,host_iface=XXXX,trsvcid=4420,hostnqn=nqn.2014-08.org.nvmexpress:uuid:f69d0000-xxxx-xxxx-xxxx-xxxxxxxxxxxx,hostid=f69d0000-xxxx-xxxx-xxxx-xxxxxxxxxxxx,discovery,ctrl_loss_tmo=600'
Failed to write to /dev/nvme-fabrics: Input/output error
failed to add controller, error failed to write to nvme-fabrics device

With the file /etc/nvme/hostid:

root@host:~# /usr/sbin/nvme connect-all --transport=tcp --traddr=XX.XXX.X.X --trsvcid=4420 --host-iface=XXXX --host-traddr=XX.XXX.X.X -n nqn.2000-02.com.mikrotik:xy_zab_raid6_array -vv
scan controller nvme0
lookup subsystem /sys/class/nvme-subsystem/nvme-subsys0/nvme0
skipping namespace scan for ctrl nvme0
skipping path scan for ctrl nvme0
scan subsystem nvme-subsys0
skipping namespace scan for subsys nqn.1994-11.com.samsung:nvme:PM9A1a:M.2:XXXXXXXXXXXXXX
kernel supports: instance cntlid transport traddr trsvcid nqn queue_size nr_io_queues reconnect_delay ctrl_loss_tmo keep_alive_tmo hostnqn host_traddr host_iface hostid duplicate_connect disable_sqflow hdr_digest data_digest nr_write_queues nr_poll_queues tos keyring tls_key fast_io_fail_tmo discovery dhchap_secret dhchap_ctrl_secret tls concat recovery_delay 
connect ctrl, 'nqn=nqn.2000-02.com.mikrotik:xy_zab_raid6_array,transport=tcp,traddr=XX.XXX.X.X,host_traddr=XX.XXX.X.X,host_iface=XXXX,trsvcid=4420,hostnqn=nqn.2014-08.org.nvmexpress:uuid:f69d0000-XXXX-XXXX-XXXX-XXXXXXXXXXXX,hostid=f69d0000-XXXX-XXXX-XXXX-XXXXXXXXXXXX,discovery,ctrl_loss_tmo=600'
Failed to write to /dev/nvme-fabrics: Input/output error
failed to add controller, error failed to write to nvme-fabrics device

And with nvme connect command (with hostid and hostnqn differing):

root@zn-hyp-inf-01:~# /usr/sbin/nvme connect --transport=tcp --traddr=XX.XXX.X.X --trsvcid=4420 --
host-iface=XXXX --host-traddr=XX.XXX.X.X -n nqn.2000-02.com.mikrotik:xy_zab_raid6_array -vv
scan controller nvme0
warning: use hostid '44a35d9d-XXXX-XXXX-XXXX-XXXXXXXXXXXX' which does not match uuid in hostnqn 'nqn.2014-08.org.nvmexpress:uuid:f69d0000-XXXX-XXXX-XXXX-XXXXXXXXXXXX'
lookup subsystem /sys/class/nvme-subsystem/nvme-subsys0/nvme0
skipping namespace scan for ctrl nvme0
skipping path scan for ctrl nvme0
scan subsystem nvme-subsys0
skipping namespace scan for subsys nqn.1994-11.com.samsung:nvme:PM9A1a:M.2:XXXXXXXXXXXXXX
warning: use hostid '44a35d9d-XXXX-XXXX-XXXX-XXXXXXXXXXXX' which does not match uuid in hostnqn 'nqn.2014-08.org.nvmexpress:uuid:f69d0000-XXXX-XXXX-XXXX-XXXXXXXXXXXX'
kernel supports: instance cntlid transport traddr trsvcid nqn queue_size nr_io_queues reconnect_delay ctrl_loss_tmo keep_alive_tmo hostnqn host_traddr host_iface hostid duplicate_connect disable_sqflow hdr_digest data_digest nr_write_queues nr_poll_queues tos keyring tls_key fast_io_fail_tmo discovery dhchap_secret dhchap_ctrl_secret tls concat recovery_delay 
connect ctrl, 'nqn=nqn.2000-02.com.mikrotik:xy_zab_raid6_array,transport=tcp,traddr=XX.XXX.X.X,host_traddr=XX.XXX.X.X,host_iface=XXXX,trsvcid=4420,hostnqn=nqn.2014-08.org.nvmexpress:uuid:f69d0000-XXXX-XXXX-XXXX-XXXXXXXXXXXX,hostid=44a35d9d-XXXX-XXXX-XXXX-XXXXXXXXXXXX,ctrl_loss_tmo=600'
connect ctrl, response 'instance=1,cntlid=2'
nvme1: nqn.2000-02.com.mikrotik:xy_zab_raid6_array connected
lookup subsystem /sys/class/nvme-subsystem/nvme-subsys0/nvme1
lookup subsystem /sys/class/nvme-subsystem/nvme-subsys1/nvme1
connecting to device: nvme1

[igaw]

• connect-all
nqn=nqn.2000-02.com.mikrotik:xy_zab_raid6_array,transport=tcp,traddr=XX.XXX.X.X,host_traddr=XX.XXX.X.X,host_iface=XXXX,trsvcid=4420,hostnqn=nqn.2014-08.org.nvmexpress:uuid:f69d0000-xxxx-xxxx-xxxx-xxxxxxxxxxxx,hostid=f69d0000-xxxx-xxxx-xxxx-xxxxxxxxxxxx,discovery,ctrl_loss_tmo=600
• connect:
  nqn=nqn.2000-02.com.mikrotik:xy_zab_raid6_array,transport=tcp,traddr=XX.XXX.X.X,host_traddr=XX.XXX.X.X,host_iface=XXXX,trsvcid=4420,hostnqn=nqn.2014-08.org.nvmexpress:uuid:f69d0000-XXXX-XXXX-XXXX-XXXXXXXXXXXX,hostid=44a35d9d-XXXX-XXXX-XXXX-XXXXXXXXXXXX,ctrl_loss_tmo=600

The connect-all tries to connect to a discovery controller ('discovery' argument). Is there one?