Replicating a session to a server joining the cluster

Question

It is my understanding that session data is replicated only during calls to HttpSession.setAttribute() and HttpSession.removeAttribute(). If a user has established a session on server A (secondary on server B) and server B goes down, then an incomplete secondary is created on some other server. If server A fails before this user accidentally causes all of the session data to be replicated, then some requests to the new managed server will work and some will not creating eratic errors.
          

Answer 1

Yes the replication behaviour in 6.x is different from 51
          "I believe that we "lazily" replicate sessions in 6.x - that is, until you
          access the session,
          it is not replicated.
          The reason for this is performance. Imagine if node1 had thousands of
          sessions when node2
          was down. If we did not do session replication lazily, then as soon as node2
          was viable, it
          would be flooded with thousands of requests to replicate the existing
          sessions. I believe this is
          the reason the change was made in the architecture between 5.1 and 6.x.
          There's always a trade-off. Some customers want performance and some
          customers want reliability.
          It looks like either we can choose the reliable behavior as same as WLS5.1
          does, or explain the "lazily"
          replication of HttpSessions in our edocs."
          If this behaviour is not acceptable for you, pls contact support and raise
          your voice.
          Also reference CR070084.
          Kumar
          "Russ Hampleman" <[email protected]> wrote in message
          news:[email protected]...
          >
          > This is the exact behavior I am seeing with Weblogic 6.1sp1. This is not
          the case
          > with Weblogic 5.1sp8. Is this a bug with 6.1?
          >
          > Ben Valentine <[email protected]> wrote:
          > >It is my understanding that session data is replicated only during calls
          > >to HttpSession.setAttribute() and HttpSession.removeAttribute(). If
          > >a user has established a session on server A (secondary on server B)
          > >and server B goes down, then an incomplete secondary is created on some
          > >other server. If server A fails before this user accidentally causes
          > >all of the session data to be replicated, then some requests to the
          > >new managed server will work and some will not creating eratic errors.
          >