Re: Defect in XBD 3.386_System_Reboot

[yyyyyyyyyy@xxxxxxx]

In a message dated 2003-08-10 1:00:16 PM Eastern Daylight Time, Larry.Dwyer@hp.com writes:

> Not on a Highly Available (HA) system.  They neither shutdown nor reboot.  No portable application can be written to deal with system transitions (such as memory replacement, CPU replacement, disk replacement, interface card replacement ...
>
> Should the POSIX standard mandate that HA systems must reboot?
>
>                 Cheers,
>                 Larry 


Not at all, but these systems do have features for reinitializing system data structures and notifying running processes of resource reavailibility or addition; a mini-reboot of the 'associating generic drivers with specific hardware' phase I listed as part of a full reboot. Generic features directly supporting specifics of this variety of system is a topic for future directions of the standard I brought up in passing last year as part of the reasoning why the conf constants should be deshaded.

Logically, I believe a new Option Group for notification events and supporting interfaces, which an application can use to trigger optional behaviors based on a sysconf() call rather than ifdefs, on a HA system can be formulated based on common features of exisiting HA implementations. At this stage of the standard, leaving it as implementation-defined allows application writers to glean those common aspects from the PCDs directly, or have the PCDs reference other system docs, to the extent appropriate for what they're trying to do. Personally, I'd prefer something hammered out that could be incorporated as a mandatory part of the standard, to support the graceful shutdown and restart of applications on systems with UPSes after a power failure detected.

Sincerely,
Mark