[Kimchi-devel] [PATCH] issue #325: Use RamSession instead of FileSession

[Aline Manera]

On 02/25/2014 12:47 PM, Shu Ming wrote:
>  2014/2/25 23:22, Adam King:
>>
>> On 2/25/2014 8:42 AM, Sheldon wrote:
>>> Not sure RamSession is a good choice.
>> What are your concerns with RamSession?
>>
>>>
>>> What's the different between MemcachedSession and RamSession.
>> Memcached is a distributed memory cache: 
>> http://en.wikipedia.org/wiki/Memcached
>> which seems like overkill given that we do not presently have a 
>> clustering solution.
>>
>> At this point we are using the session to cache information about the 
>> user:
>>     groups that they are in
>>     whether or not they have sudo
>>     others?
> The user session will totally lost When Kimchi server get reboots and 
> this is a big surprise to the user.

I don't think it is a problem.

>
>>
>> All of those are fairly simple to reconstruct in the case that the 
>> session gets lost or destroyed.
>> i.e. RamSession seems appropriate for the current session usage pattern.
> Even for RamSession, we need to check that no racing condition exists 
> when two requests access the big session dictionary at the same time.  
> I searched a bit, no clear answer yet.
>

It is already done by this patch: acquire_lock() and release_lock() 
while accessing the session data

> _______________________________________________
> Kimchi-devel mailing list
> Kimchi-devel at ovirt.org
> http://lists.ovirt.org/mailman/listinfo/kimchi-devel
>