FUSE Message Broker
  1. FUSE Message Broker
  2. MB-834

using durable subs messages stay in KahaDB indefinitely if broker is restarted

    Details

    • Similar Issues:
      Show 10 results 

      Description

      Test case from DEV-2707 comment http://fusesource.com/issues/browse/DEV-2707?focusedCommentId=64476&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#action_64476 demonstrates that after running for a while, with restarts every 3 minutes, KahaDB log files build up and seem to never be reclaimed. For instance, log files look like:

      drwxr-xr-x  23 user  staff      782 Mar  3 14:23 .
      drwxr-xr-x   4 user  staff      136 Mar  3 13:22 ..
      -rw-r--r--   1 user  staff  5243078 Mar  3 13:25 db-4.log
      -rw-r--r--   1 user  staff  5243010 Mar  3 13:26 db-5.log
      -rw-r--r--   1 user  staff  5243644 Mar  3 13:50 db-38.log
      -rw-r--r--   1 user  staff  5243478 Mar  3 13:51 db-39.log
      -rw-r--r--   1 user  staff  5243003 Mar  3 13:57 db-48.log
      -rw-r--r--   1 user  staff  5242976 Mar  3 13:59 db-51.log
      -rw-r--r--   1 user  staff  5243047 Mar  3 14:12 db-74.log
      -rw-r--r--   1 user  staff  5242897 Mar  3 14:13 db-75.log
      -rw-r--r--   1 user  staff  5242950 Mar  3 14:15 db-78.log
      -rw-r--r--   1 user  staff  5242998 Mar  3 14:15 db-79.log
      -rw-r--r--   1 user  staff  5242924 Mar  3 14:16 db-80.log
      -rw-r--r--   1 user  staff  5242965 Mar  3 14:18 db-83.log
      -rw-r--r--   1 user  staff  5243749 Mar  3 14:18 db-84.log
      -rw-r--r--   1 user  staff  5243027 Mar  3 14:19 db-85.log
      -rw-r--r--   1 user  staff  5242999 Mar  3 14:20 db-86.log
      -rw-r--r--   1 user  staff  5242904 Mar  3 14:22 db-89.log
      -rw-r--r--   1 user  staff  5243047 Mar  3 14:23 db-90.log
      -rw-r--r--   1 user  staff  4718592 Mar  3 14:23 db-91.log
      -rw-r--r--   1 user  staff  6303744 Mar  3 14:23 db.data
      -rw-r--r--   1 user  staff      553 Mar  3 14:23 db.free
      -rw-r--r--   1 user  staff  3299608 Mar  3 14:23 db.redo
      

      whereas when there are no restarts, the log files eventually do get deleted:

      drwxr-xr-x  9 user  staff      306 Mar  3 15:30 .
      drwxr-xr-x  4 user  staff      136 Mar  3 15:04 ..
      -rw-r--r--  1 user  staff  5243016 Mar  3 15:29 db-74.log
      -rw-r--r--  1 user  staff  5243353 Mar  3 15:29 db-75.log
      -rw-r--r--  1 user  staff  5243785 Mar  3 15:30 db-76.log
      -rw-r--r--  1 user  staff  4718592 Mar  3 15:30 db-77.log
      -rw-r--r--  1 user  staff  8957952 Mar  3 15:30 db.data
      -rw-r--r--  1 user  staff  3303712 Mar  3 15:30 db.redo
      -rw-r--r--  1 user  staff        0 Mar  3 15:04 lock
      

        Activity

        Hide
        Gary Tully
        added a comment -

        again, thanks for the test. I found the culprit, the durable unsubscribe processing was not correctly persisted so after a restart, stale state was in use which blocked journal data file cleanup.
        Issue: https://issues.apache.org/jira/browse/AMQ-3206

        Show
        Gary Tully
        added a comment - again, thanks for the test. I found the culprit, the durable unsubscribe processing was not correctly persisted so after a restart, stale state was in use which blocked journal data file cleanup. Issue: https://issues.apache.org/jira/browse/AMQ-3206
        Hide
        Gary Tully
        added a comment -

        fix on 5.4.x-fuse, will be in the next release and in tonights snapshot.

        Show
        Gary Tully
        added a comment - fix on 5.4.x-fuse, will be in the next release and in tonights snapshot.

          People

          • Assignee:
            Gary Tully
            Reporter:
            Jonathan Anstey
          • Votes:
            0 Vote for this issue
            Watchers:
            0 Start watching this issue

            Dates

            • Created:
              Updated:
              Resolved: