Short instruction. It do not replace full document in 'docs'. This is set of scripts to install Oracle10.2 RAC on SuSe9 SP3 - and detailed step-by-step documentation. To install Oracle RAC cluster: - Create SAN volume for OCRFiel and CSSFile, make partitions (110 MB for OCRFile and 22MB for CSSFile). - Create SAN volumes for ASM of OCFSv2. You don't need partitions on these volumes if you upgraded SLES9 Sp3 to the kernel >= 283 (for x86_64) or similar (for i386). - Assign IP and names. You will need 2 IP on public network and 1 IP on private interface for each server. Configure public and private ethernets. Don't configure vip IP to any interface. See FILES.d/*/hosts as an example. - Double check /etc/hosts file and DNS. 90% RAC installation problems comes from the wrong IP addresses or host names. - Install SLES9 SP3 build 283 on both RAC servers. It should include X11 and development tools. - unzip Oracle 10.2.0.1 database, 10.2.0.1 clusterware, patchset 10.2.0.3 into the shared storage (NFS). Optionally unzip Oracle 10.2.0.3 EM agent(s). - Mount both these scripts and Oracle installation sources using NFS on both RAC servers. - Login as a root and cd into the script(s) directory. (I recommend to create symlink as /INST for simplicity). - copy CONFIG.sh-dist into CONFIG.sh and edit it. - Read RAC installation manuals in 'docs'. These manuals can be outdated (README.txt file is up-to-date while docs are not). Now run scripts one after another on both servers - if script suffix is .sh, and on 1 server only, if suffux is .1sh For example: ./230*sh List os scripts/actions: ============================================================================ 000-check-config.sh - checks CONFIG.sh 002-setup-iscsi.sh - (optional) configure iscsi. 007-copy-setduplex.sh - (optional) installs setduplex script 010-install-orarun.sh - installs orarun rpm and creates oracle 011-copy-rawnames.sh - installs rawnames script and config. 012-change-udev-perms.sh - adjust permissions for raw devices 014-move-oracle.sh - (if OBASE != /opt) relocates oracle home 015-edit-oraprofile.sh - adjust oracle profile @ /etc/profile.d 016-set-reboot-on-panic.sh - adjust system parameters 018-edit-rcoracle.sh - replace rcoracel script(s) 019-edit-o2cb.sh - (optional) adjust OCFSv2 020-start-rcoracle.sh - starts rcoracle (set up sysctl) 025-now-reboot-everything.sh - reboot both nodes to check setting(s) 030-check-oracle-user.sh - last check of environment 080-ssh-genkeys.sh - set up pwd-less shh access 090-verify-ssh-access.sh - complete ssh setting 100-verify-rawfiles.sh - verify rawnames configuration 198-run-rootpre.sh - (optional) check if crs is not running yet 200-InstCRS.1sh - installs CRS 10.2.0.1 210-runInventory.sh - (can do it manually) first root shell 220-runRootsh.sh - (can do it manually) second root sh 225-fix-bashrc.sh - (opt)improve root's .bashrc 230-check-crs.sh - (opt)shows CRS status 250-run-vipca.1sh - runs vipca (if skipped in CRS install) 260-PatchSetCRS.1sh - Updates CRS to 10.2.0.3 270-APPLY-PATCHSET.sh - apply update to CRS on all nodes 300-check-env.sh - check env for oracle user 300README.txt 310-InstOracle.1sh - Installs DB 10.2.0.1 (you run root.sh) 400-PatchSetDB.1sh - Installs PatchSet 10.2.0.3 440-run-netca.1sh - COnfigure listener(s) 500-install-agent.sh - (opt) Installs OEM agent 510-edit-gcstartup.sh - (opt) adjust OEM agent startup 600-create-database.1sh - (opt) starts dbca 610-edit-oratab.sh - (opt) adjust /etc/oratab 620-config-hangcheck.sh - (opt) adjust hangcheck_timer ============================================================================ Possible problems: - if root.sh fail in CRS installation, create partition for OCRFile and CSSFile (change size), reconfigure rawnames.conf on BOTH servers, reboot BOTH servers and try root.sh again. Some bug prevents CRS from creating OCRFile if it's size is not well adjusted. - vipca is started manually if CRS installation refuzed to configure virtual IP (reason is not known). - Dont change script-s order. It is optimzied to simplify patchset upgrade. Installation is done as: Oracle: 10.2.0.3 for x86_64 Linux: SuSe Linux Enterprise Server 9, with ServicePack 3 upgraded to the kernel 283. Shared storage: NetApp (I Used FAS270c cluster) with iSCSI license (no NFS) Storage method in Oracle: ASM Run scripts one by one. Scripts which ends as .sh must be started on ALL nodes. Scripts which ends as 1sh must be started on node1 only. See original documentation in docs directory. After installation (including database) reboot node2, wait until everything comes up (use crs_stat -t) and then reboot node 1 and check it. Possible (known) reasons why something do not comes: - CSSFile must be owned by oracle (disk group is not enough). By default udev set up 'root' owner for the raw devices (and it may happen AFTER rawnames set up oracle user). - Database wil not start if FLASH recovery area configured and do not exists (even if it is disabled). - By default, isdn daemon is running and can sometimes freeze on shutdown. Unfortunately Linux shutdown have not timeout (so it wil freeze forever).