can you try the same process but with UDP protocol? maybe it's a TCP
problem (not a memcached problem)
i'm thinking about possible problems, if you don't want to try it...

on linux i had the same problem but with another software (network
file system, and apache) no error code was reported, but at
/var/log/messages i could get the arp table overflowed

maybe this could help you to solve the problem...
since my first problem (apache/nfs) wasn't at level3, it's at level 6,7

in my case, arp was running (it didn't stoped) but didn't accept a new
'record' in table (a new client)
after some time the cache get out (near 2 minutes!) and my program get
normal again
try with UDP... i think it can help, if not let's think about other problem

2011/2/8 Dustin <[email protected]>:
>
> Thank you for not responding in the bug.  I would like to reduce
> confusion over there.
>
> On Feb 7, 5:53 pm, Roberto Spadim <[email protected]> wrote:
>> no not below layer3, it´s a overfloow of table
>
>  ARP is a mechanism to find layer 3 addresses on layer 2.
>
>  I still don't see why you believe that an address lookup table is
> causing this exact error on exactly one platform...
>
>> on linux you will get a table overflow (cache), wait some time and you
>> will get you network running again (i don´t know how to flush arp
>> cache)
>
>  ...which is not Linux.
>
>  If you know why this exact error (which claims to not be a layer 3
> write failure with no actual error code recorded) on Windows would be
> caused by address resolution for three different users under
> relatively no traffic, then that may lead us to a solution.
>
>  I don't actually know Windows enough to know why it would happen.  I
> also don't know where these builds came from.  I'd blame the memcached
> Windows code long, long before assuming Windows can't write data over
> a socket due to ARP.



-- 
Roberto Spadim
Spadim Technology / SPAEmpresarial

Reply via email to