Open Stack

Hi,

i need recommendations to setup block storage with dell storage center
iscsi drivers.

As seen in doc
(http://docs.openstack.org/developer/openstack-ansible/mitaka/install-guide/configure-cinder.html),
no need for ISCSI block storage to have a separate host.
So, i modify env.d/cinder.yml to remove "is_metal: true", and configure
openstack_user_config.yml with:
(http://docs.openstack.org/mitaka/config-reference/block-storage/drivers/dell-storagecenter-driver.html)

storage_hosts:
  p-osinfra01:
    ip: 172.29.236.11
    container_vars:
      cinder_storage_availability_zone: Dell_SC
      cinder_default_availability_zone: Dell_SC
      cinder_default_volume_type: delliscsi
      cinder_backends:
        limit_container_types: cinder_volume
        delliscsi:
          volume_driver:
cinder.volume.drivers.dell.dell_storagecenter_iscsi.DellStorageCenterISCSIDriver
          volume_backend_name: dell_iscsi
          san_ip: 172.x.y.z
          san_login: admin
          san_password: xxxxxxxx
          iscsi_ip_address: 10.a.b.c
          dell_sc_ssn: 46247
          dell_sc_api_port: 3033
          dell_sc_server_folder: Openstack
          dell_sc_volume_folder: Openstack
          iscsi_port: 3260

Same for p-osinfra02 and p-osinfra03.

I launch playbook os-cinder-install.yml and i have 3 cinder-volume
containers each on my infra hosts.
Everything is ok.

In horizon, i can create a volume (seen on the storage center) and can
attach this volume to an instance. Perfect !

But now, if i launch an instance with "Boot from image (create a new
volume)", i got an error from nova "Block Device Mapping is Invalid".
I checked my cinder-volume.log and i see:
ERROR cinder.volume.flows.manager.create_volume
FailedISCSITargetPortalLogin: Could not login to any iSCSI portal
ERROR cinder.volume.manager ImageCopyFailure: Failed to copy image to
volume: Could not login to any iSCSI portal.

I test in one container iscsi connection:
root at p-osinfra03-cinder-volumes-container-2408e151:~# iscsiadm -m
discovery -t sendtargets -p 10.a.b.c
10.a.b.c:3260,0 iqn.2002-03.com.compellent:5000d31000b4a724
10.a.b.c:3260,0 iqn.2002-03.com.compellent:5000d31000b4a728
10.a.b.c:3260,0 iqn.2002-03.com.compellent:5000d31000b4a723
10.a.b.c:3260,0 iqn.2002-03.com.compellent:5000d31000b4a727

But when login, i got:
root at p-osinfra03-cinder-volumes-container-2408e151:~# iscsiadm -m node
-T iqn.2002-03.com.compellent:5000d31000b4a724 --login
Logging in to [iface: default, target:
iqn.2002-03.com.compellent:5000d31000b4a724, portal: 10.a.b.c,3260]
(multiple)
iscsiadm: got read error (0/0), daemon died?
iscsiadm: Could not login to [iface: default, target:
iqn.2002-03.com.compellent:5000d31000b4a724, portal: 10.a.b.c,3260].
iscsiadm: initiator reported error (18 - could not communicate to iscsid)
iscsiadm: Could not log into all portals

I found in google a bug for use of open-iscsi inside lxc-container
(https://bugs.launchpad.net/ubuntu/+source/lxc/+bug/1226855), a bug
commented by Kevin Carter (openstack-ansible core team) as a "blocking
issue" (in may 2015).

Is that bug still relevant ?
Do i need to rather deploy my cinder-volume on compute host (metal) to
solve my problem ?
Or do you have others suggestions ?

Thanks.
Regards,

-- 
Fabrice Grelaud
Université de Bordeaux