is currently readonly. We are migrating to, please continue working there on Monday 14/12. See:

  • Ate Douma's avatar
    REPO-1881 [Backport 11.2] Implement new LockManager service and replace all... · b05161a4
    Ate Douma authored
    REPO-1881 [Backport 11.2] Implement new LockManager service and replace all deprecated HippoLock and HippoLockManager usages
    This is a cherry-picked and squashed backport from REPO-1811, REPO-1873, REPO-1874 and REPO-1875:
      REPO-1811 Implement in-memory lock manager
        (cherry picked from commit dfa6f279)
      REPO-1811 To the RepositoryImpl add a Journal ConnectionHelper accessor
        Via this Journal ConnectionHelper we can access the Journal DataSource
        via the ConnectionHelperDataSourceAccessor.
        Admittedly, a bit clumsy but we need access to the Journal DataSource
        because we need to know whether we are dealing with a database *and*
        whether we have a clustered setup : Only in that case, the lock mechanism
        has to be upgraded from in memory locking to db base locking because
        needs to be cluster wide
        (cherry picked from commit 4750450f)
      REPO-1811 Support for a Database lock manager
        The Database lock manager interaction is pretty much exactly the same
        as the MemoryLockManager interaction. The biggest difference is that
        the DbLockManager creates database based locks and the MemoryLockManager
        creates MemoryLocks. Note that a DbLock is the same as the MemoryLock
        but only contains a #destroy implementation that releases the database
        The DbLockManager is still work in progress
        (cherry picked from commit e0f19fab)
      REPO-1811 Fix the LockClusterTest setup
        (still ignored but the setup failed since we do not allow SNS
        below jcr:root any more)
        (cherry picked from commit 1dba0d49)
      REPO-1811 Support database locking
        (cherry picked from commit 92c914ef)
      REPO-1811 Support background jobs in the Memory or Db LockManager
        - All the background jobs are run by a single thread
        - By default they run every 5 seconds with an initial delay of 5 seconds
        - Every background job is wrapped in a 'synchronized runnable' the synchronized
          on the LockManager instance: The reason for this is that we do not want
          background jobs to write concurrently to possibly the same records as
          other (background) jobs
        - We have the following background jobs:
          1) UnlockStoppedThreadJanitor : releases database locks for records
             that were held by a thread that is not alive any more
          2) DbResetExpiredLocksJanitor : Resets all database rows which have
             expired locks (rows with status 'RUNNING' or 'ABORT' and expirationTime
             has passed
          3) DbLockRefresher: Refreshes the expiresTime of a lock to currentTime + refreshRateSeconds
          4) LockThreadInterrupter : Interrupts the Thread that holds the lock that has been marked 'ABORT'
        (cherry picked from commit cbf14aed)
      REPO-1811 Make sure getLocks and isLocked works cluster wide in case of a database
        (cherry picked from commit 441beff4)
      REPO-1811 Fix sql statement
        (cherry picked from commit d3d25849)
      REPO-1811 Extract abstract test and add an 'abort' test
        (cherry picked from commit 5ab1e55f)
      REPO-1811 use getHoldCount() instead of protected access
        (cherry picked from commit 7027fce3)
      REPO-1811 improve feedback
        (cherry picked from commit 0bf87cb5)
      REPO-1811 use getHoldCount() instead of protected access
        (cherry picked from commit a268b5e6)
      REPO-1811 Fix abort statement
        (cherry picked from commit dac8d9cc)
      REPO-1811 Encapsulate the localLocks object
        Since localLocks is a hashmap, all access to it should be synchronized.
        This can be easier achieved by encapsulating it and whenever needed by
        other classes, return a new object containing the same locks
        (cherry picked from commit cc3e10d9)
      REPO-1811 Enhance abort tests
        (cherry picked from commit 43f17029)
      REPO-1811 use auto closable in integration tests
        (cherry picked from commit 05143edd)
      REPO-1811 Correct the expiresTime parameter
        (cherry picked from commit 62bd68c2)
      REPO-1811 Locks in status ABORT should also be refreshed
        Namely if a lock is in status ABORT and the lock is for the *CLUSTER NODE*
        on which the DbLockRefresher runs, then it means that on this cluster
        node the Thread that should be aborted did not yet abort (for example
        still busy with its job). All we can do is wait until that job finishes
        (cherry picked from commit f038ec5b)
      REPO-1811 remove unneeded auto commit
        (cherry picked from commit 591ff099)
      REPO-1811 If a Thread is already interrupted, don't interrupt again
        (cherry picked from commit 8a6b58b3)
      REPO-1811 Integration tests that confirm the correct working of the DbLockRefresher
        (cherry picked from commit a3bc5d1f)
      REPO-1811 Tests to confirm lock expiration
        Integration tests that confirm the correct working of the
        (cherry picked from commit c2284223)
      REPO-1811 only run expires db lock test in integration test
        (cherry picked from commit 79490e41)
      REPO-1811 Fix test to have flexible logger node test
        (cherry picked from commit 547542ef)
      REPO-1811 Improve exception handling and logging
        (cherry picked from commit 4fdb9278)
      REPO-1811 Improved destroy of the LockManager and other improvements
        (cherry picked from commit 589c19b8)
      REPO-1811 Remove the 'refreshRateSeconds' logic
        Just fixed 60 seconds refresh rate
        (cherry picked from commit b51f32d9)
      REPO-1811 Deprecate JCR (hippo) locking
        Note all usages of JCR Locking still need to be replaced
        (cherry picked from commit 04c17617)
      REPO-1811 Use the new LockManager instead of JCR Locking
        (cherry picked from commit 8ddbbc21)
      REPO-1811 Implement autocloseable LockResource and integration tests
        (cherry picked from commit ac0b839d)
      REPO-1811 Use auto closeable
        (cherry picked from commit 6938d49c)
      REPO-1811 Add a cleanup background job that removes rows from database of old locks
        (cherry picked from commit bb1f233b)
      REPO-1811 Implement new LockResource getters
        (cherry picked from commit e209d242)
      REPO-1811 Better tearDown
        Make sure tearDown with clearing the repository still removes
        also the database entries because if they are kept, consequent runs can fail
        (cherry picked from commit ffb7b9a6)
      REPO-1811 shorter keys
        (cherry picked from commit 87e1cb4b)
      REPO-1811 Support all databases and most specifically Oracle
        (cherry picked from commit 03072bf6)
      REPO-1811 Handle the interrupted exception gracefully
        When the lockmanager has an indication for 'abort' for the lock for the
        thread that is executing the UpdateExecutor, it means, that the current
        or a different cluster node has requested the task to be stopped. Invoking
        #cancel is the graceful stop of the job
        (cherry picked from commit 6354174b)
      REPO-1811 further refactoring and cleanup of the new (Db)LockManager
        - construction moved to LocalHippoRepository as well as further generalized
        - LockManagerFactory moved to DbLockManagerFactory and now only is responsible for constructing a DbLockManager
        - DbLockManager: added support for schemaObjectPrefix and schemaCheckEnabled, like all JR schema definitions
        - Added (handling for) new AlreadyLockedException, indicating a lock retry might be feasible
        - Merged DbHelper logic into DbLockManager (which ittself now gets passed into background Runnables) and use JR ConnectionHelper to check if table already exists
        - Added LockManagerUtilsTest, in test module, for services-api provided LockManagerUtils
        (cherry picked from commit 035c7ff9)
      REPO-1811 Improve the logging statement
        (cherry picked from commit 11f334ee)
      REPO-1811 Make sure the lockThread finishes before tearDown kicks in
        (cherry picked from commit f81fb33e)
      REPO-1811 license header
      (cherry picked from commit ab46d905)
      REPO-1811 fix unit tests
        (cherry picked from commit 84858ae4)
      REPO-1873 Support closing a LockResource by a different thread - implementation
        (cherry picked from commit 8e614323)
      REPO-1873 improve LocalHippoRepository instantiation extensibility
        (cherry picked from commit 79f365d3)
      REPO-1874 Leverage LockManager for the RepositoryScheduler replacing the deprecated HippoLock usage
        (cherry picked from commit 7e91d6c3)
      REPO-1875 add support for mssql (Microsoft SQL Server)
        Also added test configurations for mssql, postgresql and oracle in repository-test module, and updated/aligned the repository.xml and test/connection parameters for each of these.
        (cherry picked from commit 92213cda)
    (cherry picked from commit fceda60165179a1aee27e095c7c649553debd36e)
pom.xml 11.4 KB