Re: [Scst-devel] Persistent Reservations Patch for SCSI-3

[Vladislav B. <vs...@vl...>]

Ryan Manikowski, on 04/10/2010 05:13 AM wrote:
> On 4/9/2010 6:30 AM, Vladislav Bolkhovitin wrote:
>> Ryan Manikowski, on 04/09/2010 07:26 AM wrote:
>>> Vlad,
>>>
>>> Emailing you directly regarding these errors. If you would prefer that I
>>> send directly to scst-devel from now on please let me know, and I will.
>> Yes, it is better to send such questions/errors to scst-devel@, CC to me.
>>
>>> Checked out 1599 this evening and attempting to compile against kernel
>>> 2.6.18-164.15.1.el5 on Centos 5.4. The RHEL patches apply cleanly for
>>> scst_exec_req_fifo-rhel5.patch (from scst/scst/kernel/rhel/) and
>>> put_page_callback-rhel5.patch ( from
>>> scst/iscsi-scst/kernel/patches/rhel/).
>>>
>>> Everything compiles fine if I DO NOT DISABLE PROC (make enable_proc).
>>> After I disable procfs usage (make disable_proc) these are the errors
>>> received during compilation.
>> The sysfs build supports only kernels 2.6.26 and higher, because in
>> 2.6.26 internal kernel's sysfs interface had a major change, which
>> made it heavily incompatible with pre-2.6.26 version.
>>
>> I forgot to mention that in the docs. Thanks on pointing on it. Fixed.
>>
>>> Two questions before you get to the errors encountered:
>>>
>>> 1) The docs look to be a bit obsolete (specifically
>>> scst/iscsi-scst/doc/iscsi-scst-howto.txt). In one document it is stated
>>> that procfs is obsolete in favor of sysfs but there are still examples
>>> shown for using the older procfs method. It is a bit difficult to figure
>>> out whether this is ABSOLUTELY true.
>> The proc interface is obsolete, but still supported (and so far
>> default), so, I think that HOWTO should describe both methods. But,
>> perhaps, I should move the sysfs interface on the first place.
>>
>>> 2) Regarding procfs, it is mentioned that it is obsolete but the
>>> documentation fails to mention whether one can still continue to use
>>> procfs to manipulate and present targets. Can procfs still be used to
>>> manipulate and present targets, or does it have no effect whatsoever?
>> I don't understand the question. I think it's clearly written that any
>> of the interfaces can be used, but not both at the same time.
>>
> 
> Vlad,
> 
> Thank you for the reply. Your insight has been very much appreciated.
> This past week I'd been working on a proof of concept for a potential
> client that would like to implement a lower cost solution for NAS
> storage. The other proposals they are receiving are for Netapp and low
> end Dell/EMC gear.
> 
> After much trouble the system is finally up and running. It is a
> combination of Centos 5.4, drbd (for real time data replication),
> heartbeat (failover services controller) and iscst-scst (for SCSI-3
> compliant target devices). The plan is to attach the storage to a
> Windows 2008 Hyper-V cluster.
> 
> Having said that, since you've been so kind in taking the time to answer
> my questions I would like to offer some of my time to help contribute to
> the documentation and testing for the PR version of iscsi-scst. I'll
> take a look through the docs and see where I can make changes/additions
> to make things a bit easier from an end user standpoint (systems
> admins). If there are additional ways that I can get involved with the
> project, please let me know.

Ryan,

Thank you for your offer. It would be great, if you write some step by 
step HOWTO how to use iscst-scst with Hyper-V cluster as well as some 
FAQ about which troubles you went through setting it up and how you 
solved them. This info would be very useful for current and future SCST 
users.

We are currently on the "almost done" stage before 2.0, so might have 
some not yet polished issues. Therefore, don't hesitate to ask about any 
problems you have, including performance, if you are not satisfied in it.

Thanks,
Vlad