[Lustre-discuss] [ROMIO Req #940] [Fwd: Re: [Lustre-devel] a new Lustre ADIO driver]

emoly.liu Emoly.Liu at Sun.COM
Tue May 5 00:56:11 PDT 2009


Hi Rob,

I have fixed the bug found by wei-keng liao in ad_lustre_aggregate.c.
Here is the new patch(v8). If you have already applied patch_v7, you can 
use the small one(aggregate.patch) in the attachment directly.

Thanks,
LiuYing

emoly.liu wrote:
> Hi Wei-keng,
>
> Wei-keng Liao wrote:
>> Hi, LiuYing,
>>
>> I am checking ad_lustre_aggregate.c and having a problem to understand
>> the calculation of avail_cb_nodes in ADIOI_LUSTRE_Get_striping_info().
>> In line 62, I suspect this if condition may not be right:
>>     if (avail_cb_nodes == nprocs_for_coll) {
> Sorry, this condition is not right, it should be
>        if (avail_cb_nodes < stripe_count * CO) {
>
> I will update the patch soon. Thanks for your careful review.
>
> -LiuYing
>>
>> For example, when nprocs_for_coll=10, stripe_count=10, and CO=1
>> the mapping of OSTs to the I/O clients is one-to-one. Before
>> line 62, avail_cb_nodes is set to 10, but the line 62 if statement
>> changes avail_cb_nodes to 5. Is this what you intend to do?
>>
>> If you would like to keep the one I/O client per OST, shouldn't
>> avail_cb_nodes be 10 in this example?
>>
>> Wei-keng
>>
>>
>
>


-- 
Best regards,

LiuYing
System Software Engineer, Lustre Group
Sun Microsystems ( China ) Co. Limited

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: adio_driver_mpich2-1.0.7_v8.patch
URL: <http://lists.lustre.org/pipermail/lustre-discuss-lustre.org/attachments/20090505/8a22ba11/attachment.asc>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: aggregate.patch
URL: <http://lists.lustre.org/pipermail/lustre-discuss-lustre.org/attachments/20090505/8a22ba11/attachment.txt>


More information about the lustre-discuss mailing list