On 04/26/2014 01:07 AM, Mike Christie wrote:
> On 04/26/2014 01:05 AM, Mike Christie wrote:
>> On 04/25/2014 11:54 PM, Cheng Cheng wrote:
>>> Hi Mike,
>>>
>>> Thanks for replying. 
>>>
>>>> You should not have to logout the session. For Red Hat and I think
>>>> SLES, when the system is shutting down the root FS gets mounted as
>>>> read-only. iSCSI is left running. Eventually /sbin/shutdown is run and
>>>> from there the kernel is shutdown and if needed the scsi layer might
>>>> have the iscsi layer send some commands like a sync cache. So for this
>>>> you just need to leave the network up.
>>>
>>> I did some research today around this. For Debian, yes, the root FS will
>>> be remounted as read-only before shutting down/rebooting. However, the
>>> system will then stuck at sync cache, even I have left network running.
>>> I have not looked into the Linux kernel to check whether iscsi layer
>>> will send logout command to the target, but I highly suspect it won't.
>>
>> It will not logout of the target at this time. It cannot because
>> userspace is shutdown and iscsid is what sends the logout request.
>>
>> A logout is *not* needed though. It is not what is causing your problem.
>>
>> For RHEL/OEL and SUSE we do not send a logout and it works fine.
>>
>>
>>>
>>> Here is what I am going to propose:
>>>
>>>     1. I am now trying to extend iscsistart to be capable send logout
>>> command.
>>>     2. In runlevel 0 and runlevel 6, after remounting the root FS as
>>> read-only, use pivot_root and chroot to change the root to a ramfs,
>>> similar to initramfs, which is loaded with iscsistart. Then we can
>>> logout gracefully by invoking iscsistart.
>>>
>>> Please let me know what do you think of this.
>>>
>>
>> It is nice to logout but it is not going to fix your problem.
>>
>> If you are getting stuck at the sync cache stage then your network
>> probably went down. /sbin/shutdown probably shutdown the network on you.
>>
> 
> When you hit this problem, on the console around the time you see the
> sync cache sent do you then see a conn error 1011 error message?
> 
> If you reconfig the cache settings on the target so a sync cache is not
> needed do you hit this problem?
> 

As another test/workaround try deleting the scsi disks while the network
is still up. So you are running in your root ramfs, then do

echo 1 > /sys/block/sdXYZ/delete

for the iscsi disks. Then let the kernel shutdown.

-- 
You received this message because you are subscribed to the Google Groups 
"open-iscsi" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/open-iscsi.
For more options, visit https://groups.google.com/d/optout.

Reply via email to