Chapter 10. Hardware/Software Problem Determination; How To Use This Information - IBM Cluster 1350 Installation And Service

Chapter 10. Hardware/software problem determination

Contents

How to use this information

This chapter helps diagnose problems associated with the eServer Cluster 1350.
Cluster 1350 is an integrated Linux Cluster that includes IBM and Third Party
hardware and software components like server nodes and associated service
processors, storage and networking subsystems, plus Cluster Systems Management
(CSM) and General Parallel File System (GPFS) software.
Problem determination involves identifying the likely cluster component where the
problem might have occurred, and following the relevant problem determination
steps for that component.
This chapter will aid in the diagnosis of problems down to the component level.
Once a failing componenet is identified you should refer to the component's
product documentation for further actions. Links to product web sites and online
product documentation are provided in this chapter as appropriate.
Diagnosing hardware/software problems in a clustered environment requires a
basic understanding of how the components of the eServer Cluster 1350 function
together.
The cluster consists of:
v One or more 19" racks.
v From 4 to 512 Cluster Nodes. The nodes of the cluster may be an x335 or
v One Management Node (an x345) for cluster systems management and
v A Management Ethernet VLAN used for secure traffic for hardware control.
v A Cluster VLAN used for other management traffic and user traffic. Cisco
v Service Processor networks. All nodes in the cluster are connected via
v A Terminal Server network for Remote Console, using the MRV In-Reach
v A high-performance Myrinet 2000 cluster interconnect, or an additional 10/100
© Copyright IBM Corp. 2003
BladeCenter containing at least four Blade servers. The nodes are configured to
execute customer applications or provide other services required by the
customer - such as file server, network gateway, or storage server.
administration.
The Management Ethernet VLAN is used for management traffic only. It is
logically isolated for security using the VLAN capability of the Cisco Ethernet
switches, and is only accessible from the Management Node. The Cluster VLAN
and Management VLANs share the same physical Cisco switches.
switches integrated with the cluster are used for the Management Ethernet
VLAN and the Cluster Ethernet VLAN.
daisy-chained service processors (x335) and/or Remote Supervisor Adapter
cards. The first node in a daisy-chain must have a Remote Supervisor Adapter
which is Ethernet connected to the Management Ethernet VLAN.
terminal server. Optionally, the customer may elect to include an additional
network.
Ethernet.
59