Re: Use of iSCSI in domU on Oracle VM 2.1.5

  • From: "Matthew Zito" <mzito@xxxxxxxxxxx>
  • To: "Martin Bach" <development@xxxxxxxxxxxxxxxxx>
  • Date: Sun, 26 Jul 2009 15:11:19 -0700

Martin,
The kernel panic is the most useful info, as that'll show us which kernel function is crashing the instance. 

You shouldn't need to tune the scsi subsystem, it should just work.
I don't doubt that the openfiler config is correct - what I was suggesting is to set up iscsi on the openfiler box but not export/ expose/present (I don't remember the openfiler terminology) any luns. The idea was to see whether the panic was being generated by the act of creating the iscsi session or shortly thereafter, or if it was the act of attaching to the luns. More likely to be the latter, but hey, troubleshooting is troubleshooting.
I did find a xen bug having to do with iscsi initiatiors seeing multiple luns and panicking, so if you are exposing multiple luns, you might try dialing it down to 1. 

Thanks,
Matt
On Jul 26, 2009, at 10:24 AM, "Martin Bach" <development@xxxxxxxxxxxxxxxxx > wrote: 


Hi Matt,

thanks for your quick reply.
I don't get a kernel oops when performing the "iscsiadm --mode node ... 
--login" operation. /var/log/messages lists the new devices, iscsiadm
returns success and then the domU reboots. Some research I did suggests that the domU can't assign the device name(s) for the new LUNs, the error message seems to support this ("register_blkdev: cannot get major 8 for sd"). I found a FAQ which suggests turning of the SCSI subsystem in the 
kernel which is not really what I'm keen to do ...
BTW the domU will never come up again because the iscsi services try to reconnect to the same LUN and then die again, but this time with a kernel 
panic (dont't have the output available right now). The only way to
recover from that is to boot to runlevel 1 and set node.startup to manual 
instead of automatic or disable the iscsi services.
I'm not 100% sure what you mean by "disable openfiler's export of the disk 
devices" - could you please elaborate a little? I have successfully
mounted iSCSI LUNs to vmware managed virtual machines using the same setup 
so I assume the openfiler setup is correct (but I'm happy to be proved
wrong)

Thank you for your help!

Martin

>
> Do you get a kernel oops? or does the box reboot? What happens if you 
> disable the openfiler's export of the disk devices but let the domu
> instance start the iscsi session?
>
> Thanks,
> Matt
>
> --
> Matthew Zito
> Chief Scientist
> GridApp Systems
> P: 646-452-4090
> mzito@xxxxxxxxxxx
> http://www.gridapp.com
>
>
>
> -----Original Message-----
> From: oracle-l-bounce@xxxxxxxxxxxxx on behalf of Martin Bach
> Sent: Sun 7/26/2009 7:26 AM
> To: oracle-l@xxxxxxxxxxxxx
> Subject: Use of iSCSI in domU on Oracle VM 2.1.5
>
> Hello group,
>
> I am trying to use shared storage exported via iSCSI in a OEL 5.2 32bit > domU. It seems that every time I am trying to login to the target I get a 
> kernel panic. I am using Oracle VM 2.1.5, xen 3.1.4 apparently.
>
> The last entries in /var/log/messages of my domU are:
>
> Jul 26 11:44:22 crsnode1 kernel: SCSI subsystem initialized
> Jul 26 11:44:22 crsnode1 kernel: Loading iSCSI transport class v2.0-724. 
> Jul 26 11:44:22 crsnode1 kernel: iscsi: registered transport (tcp)
> Jul 26 11:44:22 crsnode1 kernel: iscsi: registered transport (iser)
> Jul 26 11:44:22 crsnode1 iscsid: iSCSI logger with pid=5412 started!
> Jul 26 11:44:22 crsnode1 kernel: scsi0 : iSCSI Initiator over TCP/IP
> Jul 26 11:44:23 crsnode1 kernel: Vendor: OPNFILER Model: VIRTUAL-DISK 
>    Rev: 0
> Jul 26 11:44:23 crsnode1 kernel:   Type:   Direct-Access
>    ANSI SCSI revision: 04
> Jul 26 11:44:23 crsnode1 kernel: Vendor: OPNFILER Model: VIRTUAL-DISK 
>    Rev: 0
> Jul 26 11:44:23 crsnode1 kernel:   Type:   Direct-Access
>    ANSI SCSI revision: 04
> Jul 26 11:44:23 crsnode1 kernel: scsi 0:0:0:0: Attached scsi generic sg0 
> type 0
> Jul 26 11:44:23 crsnode1 kernel: scsi 0:0:0:1: Attached scsi generic sg1 
> type 0
> Jul 26 11:44:23 crsnode1 kernel: register_blkdev: cannot get major 8 for 
> sd
> Jul 26 11:44:57 crsnode1 syslogd 1.4.1: restart.
>
> Straight after the "login" the machine dies.
>
> I googled for the register_blkdev... error which seems to be a common > problem with Xen, somehow the dom0 and domU get confused about device > names... Almost all documentation and newsgroup posts deal with iSCSI and 
> dom0, but no one seems to use iSCSI inside a domU.
>
> http://www.oracle.com/technology/pub/articles/wartak-rac-vm.html doesn't 
> report any problems either.
>
> Has anyone seen this? All help greatly appreciated.
>
> Best regards,
>
> Martin
>
> --
> //www.freelists.org/webpage/oracle-l
>
>
>
>

Other related posts:

  1. Home
  2. oracle-l
  3. Archive
  4. 07-2009