On 29/08/12 05:03, James Harper wrote:
>> I'm pretty sure it's not a server hardware problem (It's EEC).
>> It's only affecting a very small proportion of users and there's a clear split
>> between those who see the problem most/all of the time and those that
>> never do. RAM errors aren't that selective. RAM errors also tend to make a
>> server unstable which isn't the case.
>>
>> I still think it's probably a Windows bug, but I can't explain why when we find
>> a client with the problem, it only shows up when going via Squid and not if
>> they pull the same content directly from an origin server.
>>
> Have you done a hd+diff of a 'good' file and a 'bad' file to show it is definitely just a few bits?
>
> How reproducible is the problem?
>
> James
>
The problem is that it's very hard to reproduce. It's only happening
to a very small number of clients who are mostly too far away for us
to get physical access to the machines showing symptoms.
Andy
Received on Wed Aug 29 2012 - 13:01:52 MDT
This archive was generated by hypermail 2.2.0 : Wed Aug 29 2012 - 12:00:08 MDT