[syslinux] Domain-specific HLT when idle

[Dag Wieers]

On Sun, 3 May 2009, H. Peter Anvin wrote:

> Dag Wieers wrote:
>> On Sat, 2 May 2009, Peter Iannucci wrote:
>>
>>> In our case, the serial console is not an issue.  We would like to
>>> submit a patch to isolinux which HLTs on getchar if isolinux is not
>>> running with the serial console option.  Do you think this is the
>>> right solution to this problem?  Can we work with you to resolve this?
>>
>> I support you on this, I brought this on the table last year too. We are
>> not using serial console and this is a big issue for us too.
>>
>> In our case we do CD-based anaconda installations that (on success)
>> automatically reboot. Since we cannot eject these virtual ISO images
>> they wait on the isolinux bootprompt after this reboot.
>>
>> It may take a few days (in some cases were we do test installations
>> weeks) before the machines are booted (or re-installed) and it is
>> wasting CPU power.
>>
>> On a physical machine it's less disturbing since it is dedicated
>> hardware (still not OK for green computing) but in virtualization this
>> is a huge problem.
>
> Keep in mind that if you have control over the configuration it might be
> preferrable to set a totaltimeout which shuts down the domain...

Right, but in most cases we want someone to make a decision because we 
don't know what the next step is. Is it waiting to test the network 
patchings, to update firmware, to start the installation, to reboot into 
the newly installed system ?

Any default might be wrong in 75% of the cases, and any time waiting 
before the totaltimeout is also wasted resources. If we don't need the 
serial console, why are we wasting resources in the first place ?

I understand that for normal users syslinux is booted as an action with 
someone watching the screen. In our use-cases this is only true when the 
system/VM is powered on, after that the operational team may come back 
after some (undefined) time to perform the next step and syslinux is 
idling and burning power in the meantime :-/

PS If we had some memory to write to (VPD?) we could automate (lua ?) and 
perform the different steps (and reboots) we have to do (after a small 
timeout) and that would help, but not in all use-cases either.

-- 
--   dag wieers,  dag at wieers.com,  http://dag.wieers.com/   --
[Any errors in spelling, tact or fact are transmission errors]