1.2.4 problems with high concurrency, getting off by a request

[Kevin Watt]

I'll try to make a test program to see if I can cause it on purpose  
tomorrow :)

I logged the gets/sets from the client side, and they seemed right.   
Plus, it works perfectly under 1.2.2 (and has for 6-odd months now in  
production).

Still, I can't wait for binary mode, and hopefully not having to  
worry about irritating important newlines/whitespace/etc, as that was  
a big headache in the beginning (especially when I didn't know that  
keys couldn't have newlines/whitespace in them :) )

Ciao,
Kevin Watt
"Say nothing of my religion. It is known to God and myself alone. Its  
evidence before the world is to be sought in my life: if it has been  
honest and dutiful to society the religion which has regulated it  
cannot be a bad one." - Thomas Jefferson

On Jan 27, 2008, at 9:09 PM, dormando wrote:

> Paul Scott wrote:
>> On Sun, 2008-01-27 at 20:56 -0800, dormando wrote:
>>> First I've heard of this... I'm sure we'd love to have some data and
>>> maybe something to help reproduce this.
>>>
>>
>> I too had an issue similar to this using the PHP PECL memcache  
>> client,
>> but it was due to my own stupidity. I was running multiple sites  
>> from a
>> single memcache cluster and I forgot to differentiate the keys  
>> that were
>> being used by each app server. This resulted in each site sometimes
>> (spuriously) getting another site's data and vice versa! Duh! What a
>> brain donor!
>>
>> Anyways, I also spent some wasted time looking through and adding
>> debugging lines to the client being used only to realise that I am a
>> moron.
>
> Hmm. As described the issue sounds a little different. He's getting
> responses out of sync. Still, I'm not even sure where in the code this
> could happen. I'm more likely to blame it on a client issue, but we'll
> need data to know for sure.
>
> -Dormando