=============================================================================== Oracle Server Management Readme File Release 10.1.0.2.0 =============================================================================== 1. Changes from Release 9.2.0.1.0 2. References 3. Known Problems/limitations ================================= 1. Changes from Release 9.2.0.1.0 ================================= RAC now manages Virtual IP addresses on the cluster nodes. Before installing RAC, acquire one unused IP address per node. These VIP addresses are to be entered into the VIP Configuration Assistant run from the RAC root.sh. Do not enter IP addresses of the public or private network interfaces into VIPCA. When entering public and private names into the Oracle Universal Installer Node Entry page during CRS installation, do include the DNS domain on the public and private node names. Oracle 10g requires a shared Recovery Area. If the database files are to be stored on CFS, then the Recovery Area can also be shared via CFS. If the database files are to be stored on an Automatic Storage Management disk group, then the Recovery Area can also be shared via ASM. However, if the database files are to be stored on raw devices, then a shared directory should be configured using NFS. If 9.2 or earlier GSD is running, then stop it before installing 10g CRS by using $ORACLE_HOME/bin/gsdctl stop in the 9.2 or earlier ORACLE_HOME. SRVCTL no longer uses the Global Services Daemon (GSD) to start and stop instances. Instead it uses Cluster Ready Services (CRS). GSD is needed only to serve 9.2 and earlier clients. 10g GSD should no longer be started and stopped using $ORACLE_HOME/bin/gsdctl; instead use srvctl start/stop nodeapps. ============= 2. References ============= For detailed information on setting up the cluster database for management by EM, refer to: Oracle10g Real Application Clusters Installation and Configuration Oracle10g Real Application Clusters Administration ============================= 3. Known Problems/limitations ============================= The CLUSTER_DATABASE_INSTANCES parameter must be set to the same value on all instances. Normally you should set this parameter to the number of instances in your Real Application Clusters database. The instance startup will fail if this parameter is set to a different value from the value set in other instances. On Sun Cluster, all RAC databases on the same cluster must either be all 64-bit (eg. 10g, 9i), or all 32-bit (e.g. 9i, 8i). A mix of 32-bit RAC databases and 64-bit RAC databases on the same cluster is not supported. The growth of CRS log files in the CRS home is not limited, and can fill the disk where CRS home is located. The growth of these log files should be monitored and the log files truncated when necesary. The OCR and voting disk should be on redundant, reliable storage, for example using RAID. RAC callouts are not executed with a guaranteed ordering. They are done asynchronously, and are subject to scheduling variability. Linux: CRSD has many threads that show up as separate process IDs. This is normal. Do not change a hostname after CRS installation. This includes adding or deleting a domain qualification. According to the RAC Administrator's Guide chapter 8, the oifcfg tool for configuring network interfaces support the "storage" network interface type, but "storage" is not yet a valid network interface type in Oracle Database 10g release 1. BUG 3107894: ASM Disk discovery string must be the same on all nodes DESCRIPTION: ASM DISKS SHOULD BE ACCESSIBLE BY SAME DISCOVERY STRING ON ALL NODES WORKAROUND: When setting up raw disks on shared storage for RAC database on ASM storage, these disks should have same access path from all nodes that comprise a RAC cluster database. BUG 3235218: OCRCONFIG DOESN'T RETURN ANYTHING IF EXECUTED AS USER ORACLE IN $ORA_CRS_HOME DESCRIPTION: ocrconfig commands can only be executed by the root user, but ocrconfig does not check to ensure that the executing user is root. Also, when executing ocrconfig, if there already exists an ocrconfig.log not writable by the current user in the current directory, then ocrconfig fails without printing any error message. WORKAROUND: Become root before executing ocrconfig BUG 3311316: CLSR_START_ASM DOESN'T CHECK ENABLED/DISABLED STATUS OF ASM INSTANCE DESCRIPTION: When starting an ASM instance which has a CRS resource defined for it, the CRS resource is started even when the ASM instance has been disabled through srvctl. The CRS resource will detect that the ASM instance is disabled and shut it down. Thus, a disabled ASM instance cannot be started through SQL*Plus. WORKAROUND: Enable the ASM instance before starting it. Cluster Installation -------------------- 1. Pre-Installation Setup ------------------------- Before invoking the Oracle Universal Installer to start Oracle10g Enterprise Edition installation for CRS or RAC, perform the following steps. a. Set up the oracle user equivalence on all nodes of the cluster, including the local node. b. Log in as same OS user, with the same primary group, to install CRS and RAC. The user's primary group should be the OSDBA group. c. If the file /var/opt/oracle/srvConfig.loc exists, ensure that the user performing the installation has permissions to read it. If not, change the permissions with % chmod 644 /var/opt/oracle/srvConfig.loc The group of this file should be the dba group. d. Stop any 9.2 or earlier GSDs ---------------------------------------------------------------------------- Copyright 2004, Oracle. All rights reserved.