From: Alexei_Roudnev (Alexei_Roudnev_at_exigengroup.com)
Date: Mon Nov 05 2007 - 22:11:54 CET
Message-ID: <0f1d01c81ff0$7de0b670$7031a8c0@exigengroup.com> From: "Alexei_Roudnev" <Alexei_Roudnev@exigengroup.com> Date: Mon, 5 Nov 2007 13:11:54 -0800 Subject: Re: [suse-sles-e] iSCSI, Yast, SLES 10.1
I am not sure about Open iSCSI, but technically. when initiator discovered a
target, target reports all available access ports (interfaces)(, and then
initiator can try to establish NxM connections, where N is # of initiator
interface and M is number of target ports.
(In reall life initiators dont try initiator's interfaces, so more likely
ypou can expect M connections).
But anyway, I am 100% sure that OpeniSCSI is not well tested with pultiport
targets.
----- Original Message -----
From: "Elinor Scott-Sutter" <elinor.scottsutter@searchamerica.com>
To: "Alexei_Roudnev" <Alexei_Roudnev@exigengroup.com>; "Eberhard Moenkeberg"
<emoenke@gwdg.de>; <suse-sles-e@suse.com>
Sent: Monday, October 29, 2007 5:56 AM
Subject: RE: [suse-sles-e] iSCSI, Yast, SLES 10.1
Only one of the interfaces on the initiator side is addressed. How could it
see two paths in this case? Also, it goes on to create more and more.
Lsscsi reveals seemingly countless of the following for both sda and sdb:
[0:0:0:511] no dev FALCON IPSTOR DISK v1.0 -
I need to read the rest of the emails on this topic, but I wondered how this
would apply in a situation where only one ethnernet adapter was addressed.
Thanks.
-----Original Message-----
From: Alexei_Roudnev [mailto:Alexei_Roudnev@exigengroup.com]
Sent: Friday, October 26, 2007 7:54 PM
To: Eberhard Moenkeberg; suse-sles-e@suse.com
Subject: Re: [suse-sles-e] iSCSI, Yast, SLES 10.1
It's because SLES10 uses OpeniSCSI instead of Cisco iSCSI, and OpenISCSI
developers decided (by some badly weighed reason) do not support multiport
iSCSI access (which works pretty well in Cisco iSCSI). Your SAN system have
2 ports (2 Ethernets) for the same target, so it worked _out of the box_
with SLES9 and/or any other cisco-iscsi based linux (and with windows which
supports multiport as well), but require multipath tools (pretty
inconvenient, sorry) in SLES10.
Your 2 devices are, in reality, 2 ports to the same target and the same
device; older SCSI drivers recognized them as a single device but OpenISCSI
can not do it. Use multipath tool (and get al headache causewd by it); no
other choice (except using other linux with Cisco iSCSI driver).
PS> For us, it is primary reason why we don't use SLES10 - because it is
PS> not
compatible with multiport iSCSI targets.
Looks as a bug, but unfortunately it is a _feature_ (very bad one) so no any
hope to see it fixed.
----- Original Message -----
From: "Eberhard Moenkeberg" <emoenke@gwdg.de>
To: <suse-sles-e@suse.com>
Sent: Friday, October 26, 2007 1:07 PM
Subject: Re: [suse-sles-e] iSCSI, Yast, SLES 10.1
Hi,
On Fri, 26 Oct 2007, Elinor Scott-Sutter wrote:
> When we use Yasts iSCSI initiator module with SLES 10SP1, it appears to
> set up the initiator correctly, but in fact, I end up with two devices
> when I'm only attempting to create one.
>
> I end up with something like the following:
>
> sudo fdisk -l
>
> Disk /dev/sda: 41.9 GB, 41914728448 bytes
> 64 heads, 32 sectors/track, 39973 cylinders Units = cylinders of 2048 *
> 512 = 1048576 bytes
>
> Disk /dev/sda doesn't contain a valid partition table
>
> Disk /dev/sdb: 41.9 GB, 41914728448 bytes
> 64 heads, 32 sectors/track, 39973 cylinders Units = cylinders of 2048 *
> 512 = 1048576 bytes
>
> Disk /dev/sdb doesn't contain a valid partition table
>
> There should be no sdb. I also get for processes:
>
> root 4907 1 0 16:22 ? 00:00:00 /sbin/iscsid -c
> /etc/iscsi/iscsid.conf -p /var/run/iscsi.pid
> root 4908 1 0 16:22 ? 00:00:00 /sbin/iscsid -c
> /etc/iscsi/iscsid.conf -p /var/run/iscsi.pid
>
> I think the duplicates above might be from a logger that starts with it.
>
> Lsscsi reveals something like the following:
>
> [0:0:0:0] disk FALCON IPSTOR DISK v1.0 /dev/sda
> [0:0:0:1] no dev FALCON IPSTOR DISK v1.0 -
> [0:0:0:2] no dev FALCON IPSTOR DISK v1.0 -
> [0:0:0:3] no dev FALCON IPSTOR DISK v1.0 -
> ...
> ...
> [0:0:0:255] no dev FALCON IPSTOR DISK v1.0 -
> [0:0:0:256] disk FALCON IPSTOR DISK v1.0 /dev/sdb
> [0:0:0:257] no dev FALCON IPSTOR DISK v1.0 -
> ...
> ...
> [0:0:0:510] no dev FALCON IPSTOR DISK v1.0 -
> [0:0:0:511] no dev FALCON IPSTOR DISK v1.0 -
>
> What would cause the device to be picked up numerous times instead of
> once? It appears to create two processes and two devices when only one
> is required. And more actually.
Probably your sda and sdb are the same physical storage area.
Your IPStor host does expose your storage LUN to two of his ethernet
adapters, so you see it twice.
Use the multipath-tools to manage your multiple pathes to the single
storage device.
Viele Grüße
Eberhard Mönkeberg (emoenke@gwdg.de, em@kki.org)
--------------------------------------------------------------------------------
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: suse-sles-e-unsubscribe@suse.com
> For additional commands, e-mail: suse-sles-e-help@suse.com
---------------------------------------------------------------------
To unsubscribe, e-mail: suse-sles-e-unsubscribe@suse.com
For additional commands, e-mail: suse-sles-e-help@suse.com
---------------------------------------------------------------------
To unsubscribe, e-mail: suse-sles-e-unsubscribe@suse.com
For additional commands, e-mail: suse-sles-e-help@suse.com
---------------------------------------------------------------------
To unsubscribe, e-mail: suse-sles-e-unsubscribe@suse.com
For additional commands, e-mail: suse-sles-e-help@suse.com
This archive was generated by hypermail 2.1.7 : Tue Nov 06 2007 - 03:16:29 CET