socket options

[David Collier-Brown]

On 02/01/2013 01:12 PM, Scott Lovenberg wrote:
> On 2/1/2013 1:01 PM, Ira Cooper wrote:
>> I never denied you can shoot yourself in the foot with them.
>>
>> deprecated == going to remove at some point.
>>
>> If you have a proposed doc update with a warning about changing them,
>> and
>> their potential to have a massive negative impact on performance.  That
>> sounds quite reasonable.
>>
>> Thanks,
>>
>> -Ira
>>
>>
> I'm in favor of using wording something like what is used for "kernel
> oplocks" or one of the oplocks settings.  It basically says in the man
> page, "don't touch this unless you're sure you know what you're
> doing".  Of course, everyone _always_ knows that they know what
> they're doing, so the effect may be somewhat limited.
>
> There are still many guides out there that say to set socket options
> that these days would basically cripple your TCP/IP stack on Linux
> (and probably most other *NIXes).  People that like to "tune" will
> blindly drop these settings into their smb.conf regardless if the man
> page had a huge skull and cross bones with a mushroom cloud next to
> the option.
>
I think those people deserve the performance they get (;-))

Unfortunately, they may work for my company, so I'd love to have the
warnings, so I can diagnose their errors and correct them.   I used to
have a sysadmin whose nickname was "wooden shoes", for the amount of
self-sabotage he tended to commit. RTFM was *absolutely* necessary
around him!

--dave

-- 
David Collier-Brown,         | Always do right. This will gratify
System Programmer and Author | some people and astonish the rest
davecb at spamcop.net           |                      -- Mark Twain
(416) 223-8968