On 07/10, Oliver Wenz wrote:

Hi, I've deployed OpenStack successfully using openstack-ansible. I use cinder with LVM backend and can create volumes. However, when I attach them to an instance, they stay detached (though there's no Error Message) both using CLI and the Dashboard.

Looking for a solution I read that the cinder logs might contain relevant information but in Ussuri they don't seem to be present under /var/log/cinder...

Here's the part of my openstack_user_config.yml regarding Cinder:

storage_hosts:
   lvm-storage1:
     ip: 192.168.110.202
     container_vars:
       cinder_backends:
         lvm:
           volume_backend_name: LVM_iSCSI
           volume_driver: cinder.volume.drivers.lvm.LVMVolumeDriver
           volume_group: cinder-volumes
           iscsi_ip_address: 10.0.3.202
         limit_container_types: cinder_volume

I've created cinder-volumes with vgcreate before the installation and all cinder services are up:

# openstack volume service list +------------------+--------------------------------------+------+---------+-------+----------------------------+ | Binary | Host | Zone | Status | State | Updated At | +------------------+--------------------------------------+------+---------+-------+----------------------------+ | cinder-backup | bc1bl10 | nova | enabled | up | 2020-10-07T11:24:10.000000 | | cinder-volume | bc1bl10@lvm | nova | enabled | up | 2020-10-07T11:24:05.000000 | | cinder-scheduler | infra1-cinder-api-container-1dacc920 | nova | enabled | up | 2020-10-07T11:24:05.000000 | +------------------+--------------------------------------+------+---------+-------+----------------------------+

Hi,

Configuration option iscsi_ip_address was removed a long time ago in Cinder, the new one is target_ip_address (I don't know if the playbook maps it or what).

I recommend you run the attach request with the --debug flag to get the request id, that way you can easily track the request and see where it failed.

Then you check the logs like Dmitriy mentions and see where things failed.

It can fail on:

- cinder-volume: if it cannot map the volume (unlikely) - nova-compute: on os-brick, so you'll have a traceback

It's important that the target_ip_address can be accessed from the Nova compute using the interface for the IP defined in my_ip in nova.conf

Assuming that iscsi_ip_address is not doing anything, then the LVM driver will probably use the one defined in ip (192.168.110.202).

If my_ip is not defined in nova.conf, then you can see the default for Nova running in that compute node:

python -c 'from oslo_utils import netutils; print(netutils.get_my_ipv4())'

So make sure you can actually access from that interface the IP in Cinder.

I wouldn't bother with all that myself, I would just set debug log levels in cinder-volume and check the initialize_connection call in the logs to see the parameters in the entry call that Nova is sending (the IP it is going to be connecting from) and the return value where we can see the IP of the iSCSI target.

Hope that helps.

Cheers, Gorka.

Thanks in advance!

Kind regards, Oliver