Seeds of Genius » Errors

Live Upgrade Troubleshooting

kcassidy — Fri, 13 Nov 2009 19:39:54 +0000

This is meant to be a list of basic troubleshooting tips for Live Upgrade. It is a work in progress.

#1 and #2 Troubleshooting Tips!

Patch up to the most recent version of 121430 (SPARC) or 121431 (X86) before running any command that starts with “lu”
Incorrect syntax or an old buggy lu operation can leave the configuration in a bad state and cause lu commands to fail, even ludelete. Clean up the configuration and start fresh!
```
rm -f /etc/lutab
rm -f /etc/lu/.BE_CONFIG
rm -f /etc/lu/ICF.*
```

Helpful links:

Use this document religiously – it contains procedures, required patch levels, and known bugs: Solaris[TM] Live Upgrade Software: Patch Requirements
Update 6: Solaris 10 10/08 Installation Guide: Solaris Live Upgrade and Upgrade Planning
Update 7: Solaris 10 5/09 Installation Guide: Solaris Live Upgrade and Upgrade Planning
NGZ Supplement: How to Upgrade the Solaris 10 OS With Non-global Zones Using Solaris Live Upgrade
SVM: Examples of Using Solaris Live Upgrade to Create RAID-1 Volumes

Debugging & Helpful files: (in addition to Sun Explorer)

First, change the following lines in /etc/default/lu

LU_DEBUG=0              >>  LU_DEBUG=10
LU_SESSION_LOG_FILE=""  >>  LU_SESSION_LOG_FILE="/etc/lu/sessionlog"
LU_ERROR_LOG_FILE=""    >>  LU_ERROR_LOG_FILE="/etc/lu/errorlog"

Capture the debugging output with /usr/bin/script. Most of the output doesn’t end up in that session log file. The session and error logs are most helpful for capturing what happens during the shutdown scripts.
Grab the following:
/etc/lu/ (entire directory)
/etc/lutab
/var/sadm/system/logs/upgrade_cleanup
/var/sadm/system/logs/upgrade_log

Helpful commands:

# /etc/lib/lu/lubootdev -b
Prints the PBE’s boot device as live upgrade sees it.
# eeprom boot-device new-boot-device
SPARC – specifies a new system boot device. Without new-boot-device it prints the current one.

Error diagnosis:

luupgrade:

ERROR: Installation of the packages from this media of the media failed; pfinstall returned these diagnostics:
SunSolve Problem Resolution procedure

…work in progress.

Jumpstart of Solaris VM on VMWare ESX hangs

bwagner — Wed, 28 Oct 2009 23:02:15 +0000

Problem Description:

I have a Jumpstart infrastructure set up that has had no problems installing Solaris natively on my servers. However, when I attempt to do so within a virtual machine, the Solaris/Jumpstart installation hangs after the Grub menu. I see it get the DHCP and if I do a snoop on the interface of the jumpstart server I can see the tftp. It looks like the tftp finishes correctly and I see a couple NFS 4 requests from the client but the console on the vm never shows anything but dots and eventually the dots stop coming up. I don’t see any error messages on the client at all.

The jumpstart server is a V240 running Solaris 10. We have a /tftpboot directory for the pxegrub and inetboot files (depending on architecture). The jumpstart files are in a mount point /jumpstart where there are several different OS levels under /jumpstart/OS. The machine has multiple nics connected to various subnets.

I only have one server setup as an ESX server. On that ESX server I currently have two virtual machines, machineA and machineB. I used the DVD to install machingA since Jumpstart was not working and I set up machineB to continue testing it. I am able to ping the Jumpstart server, which is on the same subnet, from both the VMware ESX console and from the running Solaris virtual machine.

Operating System:

VMWare ESX

Troubleshooting Steps:

Verify valid GRUB bootloader entries

Root Cause:

The console= kernel parameter (e.g console=ttya) can help with switching between serial and video on physical x86 hardware, but that issue does not apply to a virtual machine.

Solution:

If such a flag is present, try removing the console=ttya flag and attempt the Jumpstart installation again.

Linux Installation: X server could not be started

bwagner — Wed, 28 Oct 2009 22:25:43 +0000

Problem Description:

During a SuSE 9 SP3 X64 installation on a Sun Fire x6250 using ILOM kvm redirection and both CD and ISO based installation media images, the graphical installer does not start with an error stating that either there is less than 96 MB of video memory (RAM) or the X Server could not start. According to Sun’s datasheet, the X6250 has an embedded AST2000 video controller with 128 MB of shared SDRAM.

Upon completing the text-based failback installation, the main OS GUI is not available.

Operating System:

SuSE Linux

Error Messages:

there is less than 96 MB memory or the X server could not be started

Troubleshooting Steps:

Using ALT+F4 to obtain a root (#) terminal, look at the X Server log by running the command:

# less /var/log/XFree86.log

Root Cause:

The XFree86.log showed X Server failed to start due to no screens found. A few lines higher showed this was because the config file had failed to parse. A few lines above that mentioned an invalid configuration parameter related to Mouse[1] and the server layout.

This problem is caused by the ILOM remote mouse mode settings. A setting of “absolute,” normally used for Windows and Solaris, will cause this problem with XFree86.

Solution:

Access the X6250 ILOM and navigate to the Redirection -> Mouse Mode settings tab. Change the mouse mode from ‘Absolute’ to ‘Relative’ and apply.

NOTE: This will reset the service processor, which will take a few moments to come back up but it should not affect the running system.