The notes in this section apply to both upgrade
types: major and minor.
Upgrade Does Not Require All Patches For Current Release |
 |
The upgrade chapter in the HP XC System Software Installation Guide (Task
1: Prepare for the Upgrade) recommends that you install all relevant
patch kits for the current release before upgrading to Version 3.1.
This step is not required.
Save the LSF Configuration and Customizations Before Upgrading |
 |
Before upgrading the current system, save the
LSF /hptc_cluster/lsf/conf/configuration directory.
Also save any other known LSF customizations, such as elim scripts.
Compare the Final LSF configuration. After the cluster_config portion
of the upgrade has completed, compare the final LSF configuration
with the saved version to ensure that only the appropriate changes
relevant to the upgrade have occurred. Also ensure (or restore, if
required) the elim scripts. HP recommends that
you use another terminal window to install the elim scripts at the point when the cluster_config utility
displays the following prompt:
All user specified configuration is complete.
The Golden Image will be created next.
[P]roceed, [Q]uit: |
Thus, changes to the /opt/hptc/lsf/top/6.2/ directory will be included in the golden image. This is the directory
where elim scripts are expected to reside.
Any changes to the LSF configuration files in
the /hptc_cluster/lsf/conf/ directory can occur
after cluster_config has finished because this
directory is not part of the golden image.
If the LSF version has changed, binary paths (including
the $LSF_SERVERDIR directory, where elim scripts are expected to reside) have also changed.
Possible Failed RPM Dependencies Errors Seen During the preupgradesys
Utility |
 |
The preupgradesys utility might report failed dependencies errors when it is trying
to remove Myrinet, InfiniBand, and QsNetII interconnect RPMs. To work around this problem, you must manually
remove the appropriate RPM.
Myrinet
Remove the gm RPM if you see
a message similar to the following:
error: Failed dependencies:
libgm.so.0()(64bit) is needed by (installed)
lustre-lite-1.4.2-2.6.9_22.11hp.3sp.XCsmp_3.1.0_SFS_2.1_1.2.x86_64 |
Use the following command to manually remove the gm RPM before proceeding to the next step in the upgrade
process.
InfiniBand
An InfiniBand RPM dependency failure looks similar
to the following:
Removing infiniband RPM and directories
error: error: Failed dependencies:
Failed dependencies:
ibhost-biz is needed by (installed) paracomp-ib-1.1-beta2.x86_64
ibhost-biz is needed by (installed) paracomp-ib-1.1-beta2.x86_64 |
Use the following command to manually remove the ibhost-biz RPM before proceeding to the next step in the
upgrade process.
# rpm -ev --nodeps ibhost-biz |
QsNetII
If you encounter a dependency failure with the
QsNetII RPM, use the following command
to manually remove the qsnetdiags RPM before proceeding
to the next step in the upgrade process.
# rpm -ev --nodeps qsnetdiags |
The upgradesys Utility Might Fail While Backing Up the Database |
 |
The upgradesys utility invokes
the managedb backup command to back up the current
database. If a database table is corrupted, the mysqldump fails with the following error message:
Backing up database to
/opt/hptc/etc/sysconfig/upgrade/upgradesys.dbbackup-20061101164854.sql ...
mysqldump: Got error: 1016: Can't open file: 'CPUTotalsLog.MYI' (errno: 145)
when using LOCK TABLES
Could not perform backup:
Bad file descriptor at /opt/hptc/database/sbin/backup line 83,
<CONF> line 12.
Can't dump databases: Bad file descriptor at /opt/hptc/sbin/upgradesys
line 127, <STDIN> line 1. |
If this failure occurs, run the following command
to correct the database tables:
# myisamchk -r /var/lig/mysql/*.MYI |
After the database tables are corrected, invoke
the upgradesys utility again.
The hptc-ire-serverlog Service Might Not Start |
 |
The hptc-ire-serverlog service is responsible for logging node configuration (nconfig)
and imaging events to the /hptc_cluster/adm/logs/imaging.log file. Follow this procedure to verify that the service is running.
If it is not running, start the service.
Use the following command
to verify that the hptc-ire-serverlog service
is running after the cluster_config [--migrate|--init] processing is complete:
# service hptc-ire-serverlog status
hptc-ire-serverlog is dead, but pid file exists |
Determine if the /hptc_cluster/adm/logs/imaging.log file exists. If this
file does not exist, use the following command to create it. If this
file does not exist, the startsys utility fails.
# touch /hptc_cluster/adm/logs/imaging.log |
If hptc-ire-serverlog is not running, start the service:
# service hptc-ire-serverlog start |