10 Jun, 2013

2 commits


07 Jun, 2013

3 commits

  • Back a few years ago, when we did the SVR4 package, we split the schema into read-only (Install) and writable (Instance).
    With the template model in 2.6, the whole schema is under config, and we should never edit/change the one in the template.
    So, I've removed all code related to the 2 locations for schema, reverted some I10N messages to their previous versions, and tidy up code and tests.
    All unit-tests are passing.
    And it's also possible to change the instance.loc and run setup and have a fully functioning server :-)
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@8978 41b1ffd8-f28e-4786-ab96-9950f0a78031
    ludo
     
  • * use single TCP packet where possible for replication messages
    * reduce frequency of socket reads.
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@8977 41b1ffd8-f28e-4786-ab96-9950f0a78031
    matthew
     
  • The jnlp / webstart is now able to display the license if it is present and to validate it.
    
    Additional fix : The font is now set to UIFactory.TEXTFIELD_FONT for the license panel. (fix windows OS trivial graphical bug)
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@8975 41b1ffd8-f28e-4786-ab96-9950f0a78031
    violette
     

06 Jun, 2013

1 commit


05 Jun, 2013

4 commits


04 Jun, 2013

2 commits

  • This commit addresses the InternalSearchMonitorTestCase unit test regression introduced by my previous fix for OPENDJ-866. 
    
    Monitor provider instance names are required to be stable so that deregistration attempts succeed. This is not the case for many of our replication monitors which include mutable components such as host addresses or connected RS identifiers (as a parent RDN). Before my fix for OPENDJ-866 these "mutating" monitors took care to re-register themselves whenever they detected that the name was about to change. Unfortunately my fix for OPENDJ-866 made the names of local replica monitors mutable as well by including the local address. The side effect was that the monitor providers would sometimes not be deregistered which was causing the org.opends.server.monitors.InternalSearchMonitorTestCase unit test to fail. 
    
    This fix modifies the replication domain monitor provider so that it re-registers itself whenever it senses a potential name change (i.e. connection establishment). The change was not as simple as I hoped because many unit tests create replication brokers with null replication domains. 
    
    I've refactored the code a little and pushed ownership of the monitor provider into the broker in order to make its lifecycle clearer.
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@8963 41b1ffd8-f28e-4786-ab96-9950f0a78031
    matthew
     
  • Align signature of DirectoryServer.deregisterMonitorProvider with registerMonitorProvider and avoid lots of boilerplate name normalization.
    
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@8958 41b1ffd8-f28e-4786-ab96-9950f0a78031
    matthew
     

03 Jun, 2013

3 commits

  • git-svn-id: https://svn.forgerock.org/opendj/trunk@8955 41b1ffd8-f28e-4786-ab96-9950f0a78031
    violette
     
  • * add host:port information to each replication node monitoring entry
    * make replication monitor entry naming more consistent
    * bump the protocol version in order to support including the remote server URL in DSInfo/TopologyMsg.
    
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@8953 41b1ffd8-f28e-4786-ab96-9950f0a78031
    matthew
     
  • - The upgrade context is now set in the CLI.
    - The upgrade context contains now the callback handler
    which simplify all the function (removed all calls to callbackhandler, functions just need the context at this point).
    - The context contains the CLI user selected options
    which simplify the code, removed the CheckCLIUserOption, the task have just to check the context for those parameters.
    - Removed the useless VerificationCallback.java
    - Added a TaskType in UpgradeTask.java.
    - Added function verifyTaskType in UpgradeTasks.java which verify user selected options are correct.
    - Modified tools.properties message numbers
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@8951 41b1ffd8-f28e-4786-ab96-9950f0a78031
    violette
     

31 May, 2013

2 commits

  • * Set the RS URL by first checking the list of configured replication servers to see if any correspond to the local host / port. If a match is found then it is used, otherwise we fall back to the old mechanism which was to perform a name lookup of /etc/hostname.
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@8950 41b1ffd8-f28e-4786-ab96-9950f0a78031
    matthew
     
  • * fix unit test regression introduced by previous commit. RS candidate filtering was too aggressive and assumed that there would only be one RS per JVM, which is not the case in the unit tests
    
    * also use a thread safe set implementation for the local RS ports in order to avoid potential race conditions in unit tests (this may be the cause of occasional RS load balancing unit test failures).
    
    
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@8948 41b1ffd8-f28e-4786-ab96-9950f0a78031
    matthew
     

30 May, 2013

3 commits

  • This partial fix addresses the first two bugs described in the issue comments, specifically:
    
    * the logic (ReplicationBroker#filterServersInSameVM) which chooses the best RS does not prefer RS which are on the same host, only those which are in the same VM. In addition, the logic in ReplicationServer#isLocalReplicationServerPort assumes that the local host name is the same address which the RS is listening on
    
    * the method ReplicationBroker#isSameReplicationServerUrl contains a typo which causes the second url to be interpreted as the local host name if the first url is a local address. This triggers all kinds of strange side effects during RS fail-over.
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@8946 41b1ffd8-f28e-4786-ab96-9950f0a78031
    matthew
     
  • - added --acceptLicense to the setup command line options.
    - modified code for checking the license during the setup :
    	- using the no-prompt option no longer bypass the license if presents.
    	- displays now the license text, question and answer(yes) when used with the --acceptLicense.
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@8943 41b1ffd8-f28e-4786-ab96-9950f0a78031
    violette
     
  • git-svn-id: https://svn.forgerock.org/opendj/trunk@8942 41b1ffd8-f28e-4786-ab96-9950f0a78031
    cjr
     

29 May, 2013

2 commits


28 May, 2013

1 commit


27 May, 2013

4 commits


24 May, 2013

1 commit


23 May, 2013

3 commits

  • git-svn-id: https://svn.forgerock.org/opendj/trunk@8903 41b1ffd8-f28e-4786-ab96-9950f0a78031
    matthew
     
  • 
    
    HTTPS was wrongly configured: code was configuring an SSLEngine that was then never used.
    Changed the code to configure an SSLEngineConfigurator that is passed down to Grizzly HTTP server.
    
    
    
    (Un)surprisingly I hit a bug on my Linux Mint:
    https://bugs.launchpad.net/ubuntu/+source/openjdk-6/+bug/1006776
    I had to modify /etc/java-6-openjdk/security/java.security so it uses the sun Elliptic Curve implementation by changing the line
        security.provider.9=sun.security.pkcs11.SunPKCS11 ${java.home}/lib/security/nss.cfg
    to
        security.provider.9=sun.security.ec.SunEC
        #security.provider.10=sun.security.pkcs11.SunPKCS11 ${java.home}/lib/security/nss.cfg
    
    
    
    HTTPConnectionHandler.java:
    Replaced instance members sslContext and sslEngine with sslEngineConfigurator.
    Renamed createSSLEngine() into createSSLEngineConfigurator().
    In createSSLContext(), removed the try / catch block duplicated with the one in createSSLEngineConfigurator().
    As a consequence, the following methods changed: configureSSL(), getEnabledSSLCipherSuites(), getEnabledSSLProtocols(), getSSLEngine(), isConfigurationAcceptable(), startHttpServer().
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@8901 41b1ffd8-f28e-4786-ab96-9950f0a78031
    JnRouvignac
     
  • OPENDJ-914 Upgrade should add Based HTTP Access Logger to config.ldif
    - HTTP connection handler and the based HTTP access logger are now added to upgrade.
    - The http-config.json is added to the config file during upgrade.
    - Simplified UpgradeTasks.java
    - Added new messages.
    - Fixed javadoc for unit test.
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@8895 41b1ffd8-f28e-4786-ab96-9950f0a78031
    violette
     

22 May, 2013

2 commits

  • OPENDJ-858 Add stats tracking to HTTP client connections
    
    
    The value output by the "cs-uri-query" extended log format was incorrect, the URL parameters should have been separated by a question mark character '?' instead of a slash character '/'.
    
    
    HTTPClientConnection.java:
    Used a question mark character '?' to build the uri query.
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@8893 41b1ffd8-f28e-4786-ab96-9950f0a78031
    JnRouvignac
     
  • 
    Protected the HTTP Connection Handler against a failing initialization: after two consecutive startup failures, the HTTP connection handler will be disabled, errors will be logged and alerts will be output.
    
    
    HTTPConnectionHandler.java:
    Added instance member enabled to decouple the enable/disable state from the config.
    Implemented getAlerts() method.
    In startHttpServer(), moved the try / catch block and the call to notify() to the run() method.
    In run(), handle failing initialization.
    In stopHttpServer(), added a null check in case of failed intitialization.
    Extracted method cleanUpHttpServer().
    
    ServerConstants.java:
    Added alerts for HTTP connection handler.
    
    protocol*.properties:
    Modified the name of the LDAP specific properties to be more generic.
    
    LDAPConnectionHandler.java
    Consequence of the change to protocol.properties.
    
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@8891 41b1ffd8-f28e-4786-ab96-9950f0a78031
    JnRouvignac
     

21 May, 2013

4 commits

  • * set the default window size to 100K
    * indicate that the properties may be deprecated in future releases
    * tag properties as advanced.
    
    Reviewed manually by Ludo.
    
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@8888 41b1ffd8-f28e-4786-ab96-9950f0a78031
    matthew
     
  • OPENDJ-858 (CR-1702) Add stats tracking to HTTP client connections
    
    
    
    In HTTP access log, changed the name of the non standard "datetime" field to "x-datetime".
    Added the "x-etime" field that outputs as a number.
    Added validation for the log format that outputs error messages.
    
    
    FileBasedHTTPAccessLogPublisherConfiguration.xml, FileBasedHTTPAccessLogPublisherConfiguration.properties:
    Renamed the "datetime" field to "x-datetime" + added the "x-etime" field.
    Improved the description of the ds-cfg-log-format.
    
    TextHTTPAccessLogPublisher.java, config.properties:
    Extracted constants for all supported field names.
    Added ALL_SUPPORTED_FIELDS constant.
    Changed logFormatFields instance member from String[] to List<String>.
    Added validateLogFormat() which can output new error messages and called it from applyConfigurationChange() and initializeLogPublisher().
    Added subtract().
    
    HTTPClientConnection.java, HTTPRequestInfo.java:
    Added instance member totalProcessingTime.
    In sendResponse(), added to the totalProcessingTime + extracted method getProcessingTime().
    Added getTotalProcessingTime().
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@8885 41b1ffd8-f28e-4786-ab96-9950f0a78031
    JnRouvignac
     
  • …d as internal operations.
    
    SdkConnectionAdapter.java:
    In close(), marked the UnbindOperationBasis as being an internal/inner operation.
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@8884 41b1ffd8-f28e-4786-ab96-9950f0a78031
    JnRouvignac
     
  • …ions when it should be
    
    Code review: matthew
    
    
    AbstractTextAccessLogPublisher.java:
    Fixed the bug by coming back a bit to what the old code was doing (it is not totally comparable).
    
    AbstractTextAccessLogPublisherTest.java:
    Updated the test.
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@8881 41b1ffd8-f28e-4786-ab96-9950f0a78031
    JnRouvignac
     

17 May, 2013

3 commits