ASGC 1

ASGC Site Status

3D Workshop @ CERN

ASGC 2

OutlinesCurrent activity

Hardware and software specifications

Configuration issues and experience sharing

ASGC 3

Current activity System migration to phase out the old system based on a mixture of

Pentium-4 and Itanium-II machines

Oracle RAC build-up to power the Oracle-related LCG services by an

extendable and more reliable Oracle system

ASGC 4

Why migration?Previous Oracle services were run by Itanium-II IA64 machines

Hardware complexity on Itanium servers made configuration quite time consuming. More standard architectures can simplify deployment.

If adopting Itanium servers, it appears that future RAC expansion will also require us to purchase Itanium servers. This is not a cost effective solution for us.

The new system is running on Pentium-D

ASGC 5

System evolution and experience in building Oracle RAC system

CPUPentium-4

3.2GHzItanium-II IA64 1.5GHz

Pentium-D 380 3.0GHz

OS SLC3 SLC4 SLC4 SLC4

Nodes(#) 2 2 2 2

ASM Config NO YES YES NO

OCFS Config YES NO NO NO

OCFS2 Config

NO YES YES YES

Past Now

ASGC 6

Hardware configuration Four servers

CPU : Intel Pentium-D 830 3.0 GHz Memory 2G (ECC) Local Disk S-ATA2 80G 7200 rpm Fiber Channel LSI 7102XP-LC, PCI

X 1

SAN Switch : Silkworm 3850 16 ports

Backend Raid subsystem: StorageTek B280

Each RAC group shares 1.7TB exported from SAN

RAC group for 3DRAC group for other LCG services

Dual channel &

redundant controller

ASGC 7

Software installationOS

SLC 4, release 2.6.9-34.0.2.EL.cernsmp

Oracle Enterprise Edition Real Application Clusters 10.2.0.1 manual patch to 10.2.0.2 (identical to CERN setup)

OCFS2 configured CRS home shared among all cluster

Partition for backup

OEM agent is under re-installation

ASGC 8

Configuration issuesHardware configuration Complexity of hierarchical SAN configuration may cause

the RAC unstable Redundant controller issues

The node will see and mount the redundant volume as the second physical volume (“rdac” module is needed to avoid this wrong behavior)

Software installation In the pre-installation stage, default OCFS kernel

modules couldn’t be loaded in SLC4 (2.6.9-34-02), recompilation from source is needed

ASGC 9

Work around on SLCBuild Oracle RAC 10g Release2 Cluster on IA64 and Scientific Linux CERN 43: http://lists.grid.sinica.edu.tw/apwiki/Build_Oracle_RAC_10g_Release2_Cluster_on_IA64_and_Scientific_Linux_CERN_43?highlight=%28rac%29 Pres-installation of Oracle RAC: http://lists.grid.sinica.edu.tw/apwiki/Pre-Installation_on_All_Nodes 3D testbed installation note: http://lists.grid.sinica.edu.tw/taskforcewiki/3D_testbed_installation%3A_oracle?highlight=%28oracle%29Oracle RAC installation note: http://lists.grid.sinica.edu.tw/taskforcewiki/Oracle_10g_RAC_Installation?highlight=%28oracle%29Oracle 10g RAC: ASM & OCFS http://lists.grid.sinica.edu.tw/taskforcewiki/Oracle_10g_RAC_related_issues%3A_ASM_%26_OCFS?highlight=%28oracle%29Oracle EM Grid Control and Agent Installation http://lists.grid.sinica.edu.tw/taskforcewiki/Oracle_EM_Grid_Control_and_Agent_Installation?highlight=%28oracle%29

ASGC 10

Extended plancomplete the installation of OEM agent

Experiment support: ATLAS / CMS

migrate Castor 2 name sever into the RAC environment

Migration of FTS backend Oracle is also considered

ASGC 11

SummaryThe Oracle services has been migrated to the new system running on 1U Pentium-D machines

A 4-node Oracle RAC has been setup with OSFS2 10.2.0.1 with manual patch to 10.2.0.2 2 nodes are dedicated for 3D Fiber channel connection to backend storage through SAN switch 1.7 TBytes storage is supported by a raid system with redundant controller OEM is under re-installation

All software configuration is based on SLC4

Site contact Yi-Ping Wu (jodiewu@gate.sinica.edu.tw) ASGC OPS (ops@lists.grid.sinica.edu.tw)