Dynamic nPartitions

[Sherif A. Louis]

Could someone please explain to me what does the "float" failure usage in the parcreate/parmodify command(s) do? Thank you.
example:
parcreate -p 0 -a 0:floating:y:float:100%

[melvyn burnard]

This sets up a cell board to "floating", or able to be removed from one partition and added to another WITHOUT bring down the partitions.
Take a read of:
http://docs.hp.com/en/10907/dynamic_nPars_WP.pdf

[Sherif A. Louis]

yes i know that. when i read the man page of the commands it stated only on failure usage option which is "ri" which means during POST if any failures occur to memory or CPU it should reboot and reactivating memory interleave. but the float option what does it exactly do? i read the white paper and it states nothing about it neither the HP-UX manpages!! how come?

[Don Morris_1]

What exactly are you wondering about it? (Because you aren't being clear here... Mr. Burnard already pretty much summed it up -- and the white paper page 4 goes into what it means (restrictions, etc.) to have a Base vs. Floating cell [so I also don't understand why you say the white paper "states nothing about it"]). Page 7 (minimum base vs. floating ratios) is important as well.

[Sherif A. Louis]

failure usage: float

i understand the floating concept cell.
i'm talking about this specific option!!

[Don Morris_1 · Accepted solution]

Ah, now I see what you're asking.

From parmodify(1m) [latest version, which I would have expected to have gone out with Dynamic nPartitions, but I don't find a patch which corresponds to it... so I'll check with some folks tomorrow and see what's going on here -- I may just not be seeing the delivery mechanism, mind you]:

failure_usage field is used by system firmware when a partition is booted.
If a CPU selftest failure or a DIMM de-allocation occurs during the power-on self-test process then this flag is used by the system firmware to determine whether or not and how the cell should be integrated into the partition at boot time.

The valid failure_usage value for cells is:

ri reactivate with memory interleave.
Specifies to integrate the cell as it would normally be integrated. This is the default for base cells.

float this is the default value for floating cells. If the user specifies ri for a floating cell, the command outputs an error and exits.

deconf deconfigure.

ni no interleave.


So the short version is -- "float" just tells the firmware to not only add the cell to the parition "ni" on hardware error, but be sure to keep in Floating. (You can't be "ri" since Floating can't be interleaved, obviously...) "deconf" could be used to keep suspect Floating cells out of the parition for checking/replacement after which they could be OLA'd.

[Sherif A. Louis]

perfect, thanks :o)

[Sherif A. Louis]

give it a '10'