|Table of Contents|
This section describes how to start and stop the NetarchiveSuite.
Note that the deploy module can make scripts for this purpose. Please refer to the [Configuration Manual 3.16] for more information on how to use the deploy module.
You need to start and stop the NetarchiveSuite applications in the correct order. The most critical part is that the BitarchiveMonitor must not start before the BitarchiveServers, as it might then initiate batch jobs before all BitarchiveServers are up and running and thus not receive the batch message. The following is a suggested order of startup:
NetarchiveSuite application startup order
- Start the databases used by NetarchiveSuite
The BitarchiveApplication (one or more) on all bitarchive servers is started:
The applications on the admin-machine are started:
- dk.netarkivet.common.webinterface.GUIApplication - dk.netarkivet.archive.arcrepository.ArcRepositoryApplication - dk.netarkivet.harvester.scheduler.HarvestJobManagerApplication - dk.netarkivet.archive.bitarchive.BitarchiveMonitorApplication for Replica One - dk.netarkivet.archive.bitarchive.BitarchiveMonitorApplication for Replica Two
- The applications on the harvester machines are started: Start each HarvesterControllerApplication instance deployed on this machine
- The applications on the access-servers are started by first starting the IndexServer and then one or more ViewerproxyApplication instances.
NetarchiveSuite application stopping order
After locating the process-id of any given process, the actually killing of the process is done on unix-machines with the kill command:
The killing itself is done in the following order:
The applications on the admin-machine are killed:
- dk.netarkivet.harvester.scheduler.HarvestJobManagerApplications. - dk.netarkivet.common.webinterface.GUIApplication - dk.netarkivet.archive.arcrepository.ArcRepositoryApplication - dk.netarkivet.archive.bitarchive.BitarchiveMonitorApplication
- Now you can shutdown the databases, if you like.
The BitarchiveApplication on all bitarchive servers are shut down:
- The applications on the harvester machines are shut down in arbitrary order:
- The applications on the access-servers are shutdown by first killing the IndexServer and then the ViewerproxyApplication instances.
Remember to empty the JMS queues after shutting down the NetarchiveSuite if you are upgrading the system or want to reset the system. If any outstanding JMS messages are around next time the NetarchiveSuite is started, they may cause deserialization errors if the message definitions have changed. To empty the JMS queue, you need to know what JMS environmentName your NetarchiveSuite instance have been using. The details of this are explained in Appendix A.
In the Danish installation, we empty the queues each time the system is restarted, so the effect of leaving messages in the queues over a restart even when not upgrading has not been tested in practice.