Uploaded image for project: 'EJB 3.0'
  1. EJB 3.0
  2. EJBTHREE-1136

Upon restart a failed node sees stale persisted SFSBs

    Details

      Description

      StatefulTreeCache needs to always purge all state in its region (including passivated sessions stored by the cache loader) as part of initialize processing. It purges as part of stop() but that's inadequate, as stop() isn't called when a node crashes.

      Effect here is if buddy replication is used and a crashed node is restarted, the buddy group tree state transfer doesn't override the old persistent state in the main tree. If a failover request then comes in for a session previously persisted in the main tree, instead of gravitating the data from the current owner the stale data is read from disk.

        Gliffy Diagrams

          Attachments

            Issue Links

              Activity

                People

                • Assignee:
                  brian.stansberry Brian Stansberry
                  Reporter:
                  brian.stansberry Brian Stansberry
                • Votes:
                  0 Vote for this issue
                  Watchers:
                  0 Start watching this issue

                  Dates

                  • Created:
                    Updated:
                    Resolved: