Specifications

www.ibm.com/support/techdocs 10

Summary of Best Practices for Storage Area Networks

• Port Statistics Counters (link reset, loss of sync, loss of signal, detection of

faulty frames, dropped frames)

• Event Logs (Fabric reconfiguration, fabric rebuild, device leaving fabric,

hardware failure)

Since ‘Event Logs’ can be configured to show (or not) different types of informational,

warning and error messages, the need to routinely clear event logs will vary. The more

different types of messages are enabled to appear in the event log, the more likely the

event log should be cleared whenever port statistics are cleared.

When port statistics are cleared, certain routine activities need to be considered. One

such example is routine maintenance activities which may cause error counters to

increment which can be problem indicators but are the result of maintenance actions.

Thus, link resets, loss of signal, loss of synchronization, CRC and other errors should

be expected to some degree whenever the following routine actions occur:

• Server or Storage leaving the fabric due to reboot, Firmware upgrade, manual

invoke of High Availability failover

• Manual invoke switch reboot, Switch Firmware upgrade

• Manual Fabric port reset

• Fabric links removed / reseated

The approach of routinely clearing of port statistics counters will insure that the statistics

represent recent time periods and not include historical events which have already been

accounted for. The time span between clearing port statistics counters is dependent on

the fabric administrator’s judgment and comfort level. The time period can range from

every couple of days to once a week to once a month. Based on the historical

experience of SAN Central, it is highly recommended the time period is not longer than

once a month.

Routine clearing port statistics should not be allowed whenever some degree of

instability in the fabric has been detected since these counters can now provide

important information about unplanned events. Once these data points have been

collected as part of the routine problem data collection routines, then it is safe to clear

the counters.

Switch names should be unique and follow some naming convention which will identify

the fabric in which the switch is a member. With modern switches now capable of being