Qlogic Fast Fabric, Руководство пользователя

Страница: 124 / 174

5 – Detailed Descriptions of Command LineTools
Health Check and Baselining Tools
5-70 D000006-000 Rev A
Q
❑
Compares present HW/SW/config to baseline
❑
Can be scheduled in hourly cron jobs
❥
As needed rerun "baseline" when expected changes occur
❑
Fabric upgrades
❑
Hardware replacements/changes
❑
Software configuration changes
❑
Etc.
5.6.2
Common Operations and Options
The Health Check and Baselining tools support the following options to select
the operations to be done by the tool:
-b
- perform a baseline snapshot of the configuration
-e
- perform an error check/health analysis only
If neither option is specified, the tool performs a snapshot of the present
configuration, compares it to the baseline and also perform an error check/health
analysis.
Use of both
-b
and
-e
on a given run is not permitted.
The typical use of the tools is to perform an initial error check by running the
-e
option. Review the errors reported in the files indicated by the tools. Once all
the errors are corrected, perform a baseline of the configuration using the
-b
option. The baseline configuration will be saved to files in
$FF_ANALYSIS_DIR/baseline
(the default of
/var/opt/iba/analysis/baseline
is set via
/etc/sysconfig/fastfabric.conf
). This baseline configuration should
be carefully reviewed to make sure it matches the intended configuration of the
cluster. If it does not, the cluster should be corrected and a new baseline run.
For example:
fabric_analysis -e
errors reported could include links with high error rates, unexpected low speeds,
etc. Correct any such errors then rerun
fabric_analysis -e
to make sure
there is a good fabric.
fabric_analysis -b
The baseline configuration will be saved to
$FF_ANALYSIS_DIR/baseline
.
This will include files starting with
links
and
comps
. These will be the results
of
iba_report -o
links and
iba_report -o
comps reports respectively.
Review these files and make sure all the expected links and components are
present. For example, make sure all the switches and servers in the cluster are