Appendix B. Frequently Asked Questions - IBM eserver Cluster 1350 Installation And Service Manual

Appendix B. Frequently asked questions

Here are some frequently asked questions about the IBM Eserver Cluster 1350.
Q: Why do I sometimes get the error message "2651-689 Java interface error for
method "query": SPException"?
A: This is due to a defect in the Remote Supervisor Adapter (RSA) firmware that is
currently being investigated by xSeries development. This problem occurs after
making between 100 and 200 connections to the RSA through the ASM library. The
work around is to reset the RSA using the web or telnet interface. Until this defect is
fixed, you may want to increase the polling interval for each hardware control point
using the command: chrsrc -s 'Name like "%"' IBM.HwCtrlPoint
PollingInterval=86400
Q: When I issue an 'rpower -n <node> reboot command why does the node not
reboot?
A: Sometimes the Remote Supervisor Adapter (RSA) cards get hung. They can be
reset via the web interface, telnetting to the RSA card, or issuing the command
rpower -n <node>resetsp_hcp.
Q: During installation process tftp hangs on the installing node. What's going on?
A: tftp is not loaded/configured on the management node.
Q: Why doesn't the xSeries 345 boot PXE correctly?
A: You cannot have a PCI ethernet card that uses the e1000 driver in the xSeries
345 when installing. Take the card out and retry the installation.
Q: Why does the dhcp server run out of leases?
A: The problem may be that you have two networks going to the same switch
fabric. This causes both eth0 and eth1 to see the dhcp requests. To fix this create
separate VLANs in the switches, one for each network attached to the switch.
Q: Why did PXE boot not get an IP address using dhcp, but the operating system
can?
A: Check the switch. All ports connected to nodes (management, compute, and
storage) should have spinning-tree turned off.
Q: Why doesn't the storage node see the drives on the FastT700, but the orange
light on the host adapter card still blinks?
A: The qla2300 driver did not load properly. Make sure the proper version of the
driver is installed.
Q: What is causing SUSE LINUX and SLES to continuously install the nodes?
A: Check that fully qualified names (host.domainname) are used in the /etc/hosts
file and that the command dnsdomainname returns the correct domain name. Also
make sure that /etc/dhcpd.conf file contains the line: 'option domain-name
"cluster.net";' Once these changes have been made run the csmsetupsiscommand
© Copyright IBM Corp. 2004
75