[Openstack-operators] Cinder with iscsi in a different IP that the cinder-api IP

Juan José Pavlik Salles jjpavlik at gmail.com
Mon Aug 12 14:56:29 UTC 2013


After reading a bit about cinder driver configuration i found out that some
conf variables are per driver. So i just added
"iscsi_ip_address=172.19.137.245" in my backends defintinions:

[lvmdriver-1]
volume_group=cinder-volumes-1
volume_driver=cinder.volume.drivers.lvm.LVMISCSIDriver
volume_backend_name=LVM_iSCSI_a
iscsi_ip_address=172.19.137.245

[lvmdriver-2]
volume_group=cinder-volumes-2
volume_driver=cinder.volume.drivers.lvm.LVMISCSIDriver
volume_backend_name=LVM_iSCSI_b
iscsi_ip_address=172.19.137.245

Works great!


2013/8/9 Juan José Pavlik Salles <jjpavlik at gmail.com>

> I tried changing in "cinder/volume/driver.py"
>
>     cfg.StrOpt('iscsi_ip_address',
>                default='$my_ip',
>                help='The port that the iSCSI daemon is listening on'),
>
> to
>
>     cfg.StrOpt('iscsi_ip_address',
>                default='172.19.137.245',
>                help='The port that the iSCSI daemon is listening on'),
>
> and it works. So for some strange reason the code is ignoring
> iscsi_ip_address in cinder.conf. Might be a bug?
>
>
> 2013/8/9 Juan José Pavlik Salles <jjpavlik at gmail.com>
>
>> Hi guys, we are using cinder with LVM+iscsi as block storage and it works
>> great so far. Now we want to change one little thing, this is what we have
>> now:
>>
>> compute-node1: eth0=172.19.136.13/24
>> storage-node1: eth0=172.19.136.245/24
>>
>> This way everything works fine, but we want to separate the cloud admin
>> network (172.19.136.0/24) from the block storage access network (
>> 172.19.137.0/24).
>>
>> We want this new config:
>>
>> compute-node1: eth0=172.19.136.13/24 eth1=172.19.137.13/24
>> storage-node1: eth0=172.19.136.245/24 eth1=172.19.137.245/24
>>
>> No problems so far, but when i attach a cinder volume to an instance the
>> iscsi connection goes through 172.19.136.13<->172.19.136.245 when i'd
>> like it to go through 172.19.137.13<->172.19.137.245.
>>
>> Looking in nova-compute.log i found:
>>
>> 2013-08-09 20:46:40.737 DEBUG nova.utils
>> [req-3d8bcf5c-9406-4681-b8e4-3a0e784098ed 20390b639d4449c18926dca5e038ec5e
>> d1e3aae242f14c488d2225dcbf1e96d6] Running cmd (subprocess): sudo
>> nova-rootwrap /etc/nova/rootwrap.conf iscsiadm -m node -T
>> iqn.2010-10.org.openstack:volume-a090f312-f419-40c3-ba86-1017a82e0968 -p
>> 172.19.136.245:3260 --op update -n node.startup -v automatic execute
>> /usr/lib/python2.7/dist-packages/nova/utils.py:208
>>
>> That explains why the connection goes that way, and maybe there's a place
>> where i can tell cinder what interface i want it to expose. It seems that
>> it's using the same IP as the API.
>>
>> I tried using
>>
>> iscsi_ip_address=172.19.137.245
>>
>> in cinder.conf but didn't work.
>>
>> Is it possible?
>>
>> Thanks!
>>
>> --
>> Pavlik Salles Juan José
>>
>
>
>
> --
> Pavlik Salles Juan José
>



-- 
Pavlik Salles Juan José
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20130812/e2fc230e/attachment.html>


More information about the OpenStack-operators mailing list