BladeSystem Management Software

Blade "power on delay" - any best practice to prevent a power spike?

 
chuckk281
Trusted Contributor

Blade "power on delay" - any best practice to prevent a power spike?

Anton had some blade power-on questions:

 

********************

 

Are you aware of any best practices or recommendations for “power-on delay” on the c-class blades?

 

I can think of 2 cases when setting power-on delay may be required:

1)      Fully populated enclosure with the servers having the automatic power-on enabled – to start blades few at a time to prevent a power spike.

2)      Customer environment specific i.e. web server connecting to a DB server which required the DB to be loaded when the app tries to connect to it i.e. delay the app to be sure the DB is up before that(same for external storage etc.)

 

Is there any possibility of a power spike or the OA will use an algorithm to prevent this?

I understand with Power capping enabled it will automatically increase caps of the individual blades, but for “default”  full config is it recommended to have power on delay enabled?

 

I guess if it is recommended HP will just implement it in the OA firmware to have delay by default on groups of bays etc but just want to doublecheck if there is best practice or anything else…(i.e recommended for single-phase AC power or so)

 

*********************

 

Monty replied:

 

********************

 

c-Class server blades since original launch in 2006 have a built-in bay based algorithm to delay the initial server power requests – for the use case of cold power-on of the entire enclosure..

 

The OA power delay feature allows the OA administrator to design a more controlled power sequence for that use case – for instance to allow the interconnects time to complete their power-on tests before allowing any of the servers to power-on, or to power-on particular servers providing shared services before other servers.

 

****************

 

Any best practices you would like to share?