[ZODB-Dev] ZEO 2 : cache twiddling

[Shane Hathaway]

Chris Withers wrote:
> Chris Withers wrote:
> 
>> Guido van Rossum wrote:
>>
>>> Oops.  All thopse unknown codes look like sometehing is seriously out
>>> of whack...
>>>
>>> I don't know what's going on -- is this on Windows?  Do you want to
>>> send me the trace file?
>>
>>
>>
>> This is on Linux. Zope 2.6.0, ZEO 2.0.
>>
>> I'll mail you the trace off-list for privacy reasons.
> 
> 
> Okay, now bumped the caches for the 3 storages used on these clients up 
> to 100Mb each. Intitial observations:
> 
> The first time I do something that involves reading from one of the 
> mounted storages, I have to wait an _age_. What's happening? Is the 
> ClientCache filling itself with all objects it can before it does 
> anything? Certainly much slower than before.
> 
> Any comments?

Uh-oh.  I wonder if you're accidentally sharing one cache between three 
storages.  This is bound to break very badly, and there's currently 
nothing that detects this condition.  Could you "ls -l" the directory 
where your cache files are?  It's usually the same directory that holds 
Data.fs.

Of course, I would expect ZODB to shatter in a million pieces if this 
were the case, rather than chug along with some occasional breakage.  Hmph.

Shane