DbConnectionBroker changes:

Release 1.0.13: 3/21/02

• Major change: Added timeout handling for connections that have been checked out of the pool. Previous versions did no timing or analysis on connections that are registered as being checked out and connections could be checked out for any amount of time. Under this scenario, connections that were not returned due to coding errors or due to database locking could exhaust the pool. In this version, the time that a connection is checked out can be monitored and a maxCheckoutSeconds parameter set to force a connection to be recycled in the event that it is not returned within this time. This feature can be turned off by supplying a zero value for the timeout parameter. The default is 60 seconds.
  SQL Server has been known to cause locks on connections in certain circumstances preventing them from being properly returned to the pool. Setting an appropriate timeout (for the database query) on connections will prevent these types of lockups from exhausting the pool.

• Also added debug levels for log reporting:
  
  • debugLevel 0: basic logging
  • debugLevel 1: Errors only
  • debugLevel 2: Errors and Warnings
  • debugLevel 3: Errors, Warnings and Informational messages

Release 1.0.12: 2/29/02

• Added ability to append to log file. This is in response to numerous complaints on the overwriting of log info making it difficult to track down reported errors.

• Bug Fix: For some databases (Sybase ASE 12.5 in particular), closed connections throw errors that prevented connections from being recycled. For these databaes, it is possible to close connections from outside an application and leave orphaned connections in the pool. This was fixed by ignoring the error on close and forcing the recycle of the connection.
  Thanks to Basheer Dargham for this fix.

Release 1.0.11: 12/7/99

• Bug Fix: Round Robin distribution of connections had a bug that caused an infinite loop if a connection timed out or the DB was shut down. (This was confirmed on mysql server and may not occur on with all databases). The result was a runaway JVM.

  Thanks to Robin Hosie for this fix.

Release 1.0.10: 12/3/99

• Bug Fix: Connection counter was incorrectly being incremented for connections having SQLExceptions. This caused the housekeeping thread to attempt to access one more connection than actually existed. This bug would cause things to lock up when the DB was brought down for any amount of time and prevented recovery on some systems. Thanks to Troy Harrison and Don Eddleman for this fix.
• Bug Fix: Connections were not being recycled correctly for some databases because the createStatement used to check health of connections was returning statement objects even when the database was shut down. Thanks to John Newlin for this fix.

Release 1.0.9: 11/24/99

• Added new destroy method allowing for controlled (Multi-phase) shutdown. This new method can be used to coordinate fail-over database implementations. Previously, the destroy method would kill all ongoing database processes indiscriminately. The new destroy method provides the following sequence:
  destroy(int millis)
  1. getConnection() will refuse to return connections.
  2. The housekeeping thread is shut down.
     Up to the time of millis milliseconds after shutdown of the housekeeping thread, freeConnection() can still be called to return used connections.
  3. After millis milliseconds after the shutdown of the housekeeping thread, all connections in the pool are closed.
  4. If any connections were in use while being closed then a SQLException is thrown.
  5. The log is closed.
  
  This method can be called from a servlet destroy() method.

• Added several other useful methods:
  • getSize()
  • getUseCount()
    
  (See docs for descriptions.)

• Bug fix: Fixed problem where log file wasn't being closed.
  Enhancement: Replaced all occurrences of Date objects with the much more efficient System.currentTimeMillis().
  Enhancement: Cleaned up run thread handling.
  

Release 1.0.8: 7/28/99

• Numerous deprecation changes. Added the feature to automatically clean up orphaned Broker Objects by trackng the process ID -- some webservers (Apache with JServ) can orphan connection pool objects as a result of the servlet reload mechanism.

Release 1.0.7: 9/19/98

• Bug fix for explicit closing of expired and/or recycled connections.  On some implementions of JDBC, connections that were restarted (do to the normal recycle expiration time or because they became unstable) were not being cleaned up by garbage collection.  This resulted in a buildup of connections on the database server.   This problem was seen with versions of JDK more recent than 1.1.4.  The fix explicitly closes the connections that are no longer in use.  Thanks to Dirk Leas and Jeff Dripps for his help with this change.

Release 1.0.6: 5/4/98

• Bug fix for logic error in getConnection() method.  For long running queries under high load, the loop logic incorrectly started up all connections (up to the maximum limit) and locked itself out of the loop.  This bug may have resulted in other anomalies in the distribution of connections.  Credit goes to Jason T. Winders for finding and fixing this bug.
• Warnings on connections are now cleared after being captured and reported to the log file.  This problem was noticed mainly in systems using the JDBC-ODBC bridge drivers and resulted in warnings being repeatedly sent to the log file.  Special thanks to Terry Billany for his input on this.

Release 1.0.5: 4/7/98

• Modified Constructor Method to handle Database Reboot timing issues.  Where DbConnectionBroker previously attempted to build the connection set once at startup, it now tries up to 20 times with a 15 second wait period.  This allows up to 5 minutes for the database to reboot or recover from a locked condition.  This was a particular problem for systems where the server and database were on the same machine -- the Servlet Server would often restart before the database causing a lockup on the Servlet Server.   This change  resolves this issue.
  Thanks to Tim McCune  for this suggestion.

Release 1.0.4: 3/28/98

• Added destroy method.  This method should be called from the destroy method of the Servlet (see the examples on the main page).  This is important for those Servers that dynamically reload servlets (e.g. JWS and ServletExpress). 
  Thanks to Dirk Leas for this recommendation.
• Enhanced output for error/log file.

Release 1.0.3: 1/10/98

• Major enhancement of housekeeping: Connections are checked more thoroughly for errors and are restarted immediately instead of the periodic reset periods. You will now be able to shut down the database and recover all connections within 10 seconds. Also, connections are actually tested for integrity instead of relying on JDBC warnings which are not dependable. If a connections becomes unusable for any reason, it is restarted immediately.
• SQLException errors are caught and logged for creating/restarting of new connections.

Release 1.0.2: 12/07/97

• Default log file created when invalid or null log filename specified. Default log file stored in local directory under name of DCB_xxxxxxxxxx.log (where the xs represent the current timestamp integer).
• Made idOfConnection method public (was private).
• Enhanced output for error/event log.

Release 1.0.1: 12/02/97

• Log file now specified by Absolute file name.
• JDBC Driver information was not being passed correctly to constructor method.

Release 1.0.0: 12/01/97

• Original release.