cubicweb #1381328 [sessions] have persistent web sessions [in-progress]

that would avoid sticky cookies hacks when behind a load balancer (and provide vastly smoother balancing)

priorityimportant
typeenhancement
done in<not specified>
load5.000
load left1.000
closed by<not specified>
patch[connection] remove a dead counter [applied][session] drop .is_repo_in_memory [applied][session] drop session-as-a-context-manager [applied][session] retire session.vreg [applied][session,repo] remove last trace of "cnxprops" [applied][devtools/repotest] use the RepoAccess object [applied][session, repository] provide repo.open_session and move .close reponsibility to session object [applied][repository] remove unused check_session method and its test [applied][web/sessions] the session managers are definitely not components [applied][web] the AuthenticationMAnager is no more an appobject [applied][web] move AbstractAuthManager near its immediate concrete subclass [applied][web] move abstract session manager to web/sessions [applied][session] remove the obsolete timestamp infrastructure [rejected]Use the persistent sessions by default [rejected][schema,session] provide a database CWSession object [rejected][cache,sessions] session persisted through a memcached server [rejected][views/auth] replace a nasty call to locals() with a flag variable [rejected][web/request] use a picklable Counter object for tab index counters [rejected]

Workflow history

from state (1)to statecommentdateUser
openin-progress2014/02/12 18:16 UTC 

Comments

  • 2010/11/15 09:17

    required for customer project.

  • 2014/04/29 15:33, written by alutz

    this could have a rather important performance impact, has that impact been studied ? Is there an easy way to plug this session data into something else that postgresql ?

    • 2014/04/29 15:35, written by alutz

      oh, and by the way, this feature would be great news for scaling cubicweb

      • 2014/04/29 16:33, written by acampeas

        Of course, that's the whole point :)

      • 2014/04/30 19:33, written by cdevienne

        yep, this is why we (unlish) need it

    • 2014/04/29 16:33, written by acampeas

      The perf. impact is not really important right now as we are seeking correctness. Hence, multiplying possible backends right now is not an immediate priority.

      Note however that there exists a second patch stack providing a caching infrastructure for, amongst other things, session (usgin e.g. a memcached server).

add comment

Ticket #1381328 - latest update on 2014/06/06, created on 2010/11/10 by Aurelien Campeas