| Summary | When upgrading an Orbix 6.3 domain to the latest Service Pack it is recommended that you bring down the domain and then install the service pack. This will not overwrite the etc or var directory and hence all user specific information will not be modified. When the domain is restarted the upgrade will be complete. However it is not always possible to bring down an Orbix domain. This article explains how an upgrade could be completed with minimal interruption to services. Note, this is not a standalone approach and has only been tested on one platform. |
|---|---|
| Article Number | 24938 |
| Environment | Tested on Orbix 6.3.3 to 6.3.5 on Solaris 10 |
| Question/Problem Description | It may not be possible to stop a domain when upgrading to the latest service pack. The steps outlined in this article will provide a way of upgrading to the latest service pack with minimal interuption to services. |
| Clarifying Information | This article only covers the migration of the locator database. The CFR and naming service may be migrated in a similar manner, but this depends on how your domain is configured. Before attempting this in production it is suggested to contact support to ensure everything is configured correctly. |
| Error Message | |
| Defect/Enhancement Number | |
| Cause | |
| Resolution | First of all stop the slave services and then ensure the locator databases are ready to be backed up: itadmin pss_db checkpoint var/nwtest/dbs/locator_master/env.ior
itadmin pss_db checkpoint var/nwtest/dbs/locator_master_priv/env.ior
itadmin pss_db pre_backup var/nwtest/dbs/locator_master/env.ior
itadmin pss_db pre_backup var/nwtest/dbs/locator_master_priv/env.ior
mkdir temp
cp -R var/nwtest/dbs/locator_master temp
cp -R var/nwtest/dbs/locator_master_priv temp
cp -R /net/suzi/local1/users/mamurphy/633domain/temp/* ./var/nwtest/dbs/ Start the new master and slave. Modify the /etc/hosts or DNS of each client, node daemon and additional service to point to the new master/slave #oldmaster
#10.65.5.30 master
#oldslave
#10.65.5.26 slave
#newmaster
10.65.4.45 master
#newslave
10.65.5.19 slave
Stop the old master services. Any clients that you have running should end up trying to contact the old master and then failing over to the new slave, which in turn will be redirected to the new master: 13:27:02 9/29/2011 [_it_orb_id_1@esugar/10.2.4.55] (IT_IIOP:8) W - failed to create connection to IPAddressImplBase[inet_address=master/10.65.5.30,port=3075,protocol=TCP] - exception: connection creation failure: java.lang.reflect.InvocationTargetException
Invocation [203], invoked on : orb32
|
| Workaround | |
| Notes | |
| Attachment |
| Created date: | 29 September 2011 |
|---|---|
| Last Modified: | 14 February 2013 |
| Last Published: | 29 September 2011 |
| First Published date: | 29 September 2011 |
| Internal Comments: |
Only publishing this internally since we ideally want customers to use the system tested approach of stopping the services.
This was tested with the old master and slave running on suzi and quatro and the new master and slave running on pdsolar10 and irsol1. The client and replicated services were ran on the Linux machine esugar. |
#KnowledgeDocs
#Orbix
