[Pvfs2-users] PVFS v2.8.0 compile problems

[Phil Carns]

Hi Tony,

The intention of the mmap-racache was to perform read ahead for files 
that are mmapped.  An example use case would be running an executable 
off of PVFS.  In that case, read ahead would be more efficient than 
waiting for pages to be read in individually as needed.

In theory it shouldn't have any impact on iozone writes.

-Phil

Tony Kew wrote:
> Dear Phil,
> 
> PVFS 2.8.1 works (insofar as I can tell) with the --enable-mmap-racache
> configure option with the following patch:
> 
> --- src/apps/kernel/linux/pvfs2-client-core.c.orig      2009-02-27 
> 15:53:50.000000000 -0500
> +++ src/apps/kernel/linux/pvfs2-client-core.c   2009-02-27 
> 15:54:22.000000000 -0500
> @@ -1609,7 +1609,7 @@ static PVFS_error post_io_readahead_requ
>         &vfs_request->in_upcall.credentials,
>         &vfs_request->response.io,
>         vfs_request->in_upcall.req.io.io_type,
> -        &vfs_request->op_id, (void *)vfs_request);
> +        &vfs_request->op_id, vfs_request->hints, (void *)vfs_request);
> 
>     if (ret < 0)
>     {
> 
> 
> I am, though, getting poor iozone performance numbers for initial write...
> I'm going to run some more iozone tests & try without the
> --enable-mmap-racache  and let you know what I find...
> 
> 
> Thanks,
> Tony
> 
> 
> Tony Kew
> SAN Administrator
> The Center for Computational Research
> New York State Center of Excellence
> in Bioinformatics & Life Sciences
> 701 Ellicott Street, Buffalo, NY 14203
> 
> CoE Office: (716) 881-8930          Fax: (716) 849-6656
> CSE Office: (716) 645-3797 x2174
>      Cell: (716) 560-0910
> 
> "I love deadlines, I love the whooshing noise they make as they go by."
>                                                          Douglas Adams
> 
> 
> 
> Tony Kew wrote:
>> Dear Phil,
>>
>> Thanks for the info, the option worked for me with 7.2.1 codebase, for
>> what its worth.  The 7.8.0 code with my patch works so far (with the very
>> limited tests I have done.)  I'll be testing over several nodes this 
>> afternoon,
>> or tomorrow.
>>
>> [...]
>>
>>
>> Phil Carns wrote:
>>> Hi Tony,
>>>
>>> I just wanted to mention that the second compile problem that you 
>>> pointed out is from a code path that gets enabled with the 
>>> --enable-mmap-racache option.  That particular option is experimental 
>>> and (as you have found) not well tested.  I would not advise using it 
>>> in a production setting.
>>>
>>> -Phil