Installation guide

Managing MTBF Statistics

HP-UX version 11.00.03 Administering Fault Tolerant Hardware 5-29

hw_path is the hardware path of the device.

If the device does not change to CLAIMED, call the CAC for further assistance. For

more information about contacting the CAC, see the Preface of this manual.

Managing MTBF Statistics

The system maintains statistics on the mean time between failures (MTBF) for each

hardware device in the system. The following sections describe how the MTBF is

calculated; how to display, clear, and set the MTBF threshold; and how to

configure the minimum number of samples, as well as two other important

variables, numsamp, and the soft error weightage, soft_wt.

For more information about the hard and soft errors that trigger the system to

evaluate the MTBF, see “Error Detection and Handling.”

MTBF Calculation and Affects

For each error that occurs, the system performs certain calculations.

If the error is a hard error, the system records the time of the error and increments

the total error count. Then the system takes the device out of service and places it

in the ERROR state. Finally, the system calculates the MTBF

and compares it with

the threshold. One of the following occurs:

■ If the MTBF is less than the threshold, the system leaves the device in the

ERROR state.

■ If the MTBF is greater than the threshold, the system attempts to enable the

device and return it to the CLAIMED state.

If the error is a soft error, the system increments the soft error count and compares

the soft error count to the soft_wt variable. One of the following occurs:

■ If the soft error count is less than the soft_wt variable, the system takes no

further action and continues to monitor the device for errors.

■ If the soft error count equals the soft_wt variable, the system records the

time of the error, increments the total error count, and clears the soft error

count. Then the system calculates the MTBF and compares it with the

threshold. One of the following occurs:

The system does not calculate MTBF until the total error count equals the numsamp variable, and

then it uses the recorded times of the last numsamp errors to calculate MTBF. If MTBF has not

yet been calculated, the system considers the MTBF value unreliable and acts as if MTBF is

greater than the threshold.