[Building Sakai] request for review: KNL-498, get last event time for a user

[Nuno Fernandes]

>
> For the cases where a user doesn't actually logout, ie just closes their
> browser, I believe the process that cleans old sessions also fires a logout
> event as well so that should be fine.
>
This is curious:
Just add the case (if needed) that there may exist more logout than login
events :)

Quite strange, isn't it? I have just checked this with our installation:
#login events:  2198475
#logout events: 2335554

Not sure if this was due to some old, already resolved issue...

Hope it helps,
Nuno



>
> Thanks for the input all.
>
> cheers,
> Steve
>
>
> On 18/05/2010, at 6:37 PM, Matthew Buckett wrote:
>
> > On 18 May 2010 09:16, Nuno Fernandes <nuno at ufp.edu.pt> wrote:
> >> +1 patch looks good!
> >> You should check the query on a Sakai db with millions of entries in
> >> SAKAI_EVENT and SAKAI_SESSION (typical on production environments) as it
> may
> >> have some performance impact !
> >
> > As an alternative implementation you could have an event listener
> > which keeps a cache (map) in memory of userId to date. When a new
> > event is seen you update the date in the cache for the user.
> >
> > This should work as events propagate across the cluster and you don't
> > need to use the DB at startup as nobody is logged in. The only issue
> > is when adding a node to the cluster, as you need to populate the
> > initial cache for that node. If we had a shared cache (aka MemCache)
> > this wouldn't be a problem.
> >
> >> Great feature!
> >
> > Indeed.
> >
> > --
> >  Matthew Buckett
> >  VLE Developer, LTG, Oxford University Computing Services
>
>


-- 
Nuno Fernandes  .  { Analyst/Programmer }

|| web  . { http://www.ufp.pt  |  http://tinyurl.com/nfgrilo  |  follow_me @
nfgrilo }
|| work . { Universidade Fernando Pessoa  |  Praça 9 de Abril, 349  |
 4249-004 Porto }
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://collab.sakaiproject.org/pipermail/sakai-dev/attachments/20100518/99c31e55/attachment.html