Qlogic Fast Fabric User Manual page 125

Q
present. Also verify the appropriate links between servers and switches are
present. If the fabric is not correctly configured, correct the configuration and
rerun the baseline.
Once a good baseline has been established, use the tools to compare the present
fabric against the baseline and check its health. For example:
fabric_analysis
Will check the present fabric links and components against the previous baseline.
If there have been changes, it will report a failure and indicate which files hold
the resulting snapshot and differences. It will also check the PMA error counters
and link speeds for the fabric (similar to fabric_analysis -e). If either of
these checks fail, it will return a non-zero exit status, therefore permitting higher
level scripts to detect a failed condition.
The differences files are generated using the Linux command specified by
FF_DIFF_CMD in fastfabric.conf. By default this is the diff -C 1
command. It is run against the baseline and new snapshot. Therefore, lines
after each *** #,# **** heading in the diff are from the baseline and lines
after each --- #,# ---- heading are from the new snapshot. If FF_DIFF_CMD
is simpy set to diff, lines indicated by "<" in the diff would be from the baseline
and lines indicated by ">" in the diff would be from the new snapshot. Another
command which can be useful is the Linux sdiff command. For more
information about the diff output format, consult the Linux man page for diff.
If the configuration is intentionally changed, a new error analysis and baseline
should be obtained using the same sequence as for the initial installation
(discussed above), establishing a new baseline for future comparisons.
In addition all of the tools support the following two options:
-s - save history of failures.
-d dir - top level directory for saving baseline, snapshots and history (default
is $FF_ANALYSIS_DIR which is set in fastfabric.conf).
When the -s option is used, each failed run will also create a directory (whose
name is the date/time the analysis tool was started) containing the failing
snapshot information and diffs. This will permit a history of failures to be
tracked. Note that every run of the tools also creates a latest directory with
the latest snapshot. The latest files are overwritten by each subsequent run
of the tool, which means the most recent run results are always available.
Beware, frequent use of the health check tools in conjunction with -s can
consume a large amount of disk space. The space requirements will depend
greatly on the size of the cluster, for example it could be > 10 megabytes per run
on a 1000 node cluster.
The -d option allows command line control over the baseline, snapshot and
history directory tree. Runs using -d must use the same directory as any
D000006-000 Rev A
5 – Detailed Descriptions of Command LineTools
5-71