[gentoo-user] Re: 2 months into an 8-month computation.

2019-07-13 Thread Grant Edwards
On 2019-07-12, Nikos Chantziaras  wrote:
> On 11/07/2019 20:59, Alan Grimes wrote:
>> 'ey, I have the 2.3 months into an 8-month computation blues...
>> [...]
>> So basically all gentoo updates will have to be done at the end of this 
>> run, I'm not really sure when, sometime in the December-January timeframe.
>
> I guess you should have written your code in a way that can store 
> current state so that it can resume.

No kidding.  Isn't "how to use checkpoint files" lesson number zero
when you start working on long-running computational jobs?

--
Grant




Re: [gentoo-user] Re: 2 months into an 8-month computation.

2019-07-12 Thread Laurence Perkins
On Fri, 2019-07-12 at 07:18 +0300, Nikos Chantziaras wrote:
> On 11/07/2019 20:59, Alan Grimes wrote:
> > 'ey, I have the 2.3 months into an 8-month computation blues...
> > [...]
> > So basically all gentoo updates will have to be done at the end of
> > this 
> > run, I'm not really sure when, sometime in the December-January
> > timeframe.
> 
> I guess you should have written your code in a way that can store 
> current state so that it can resume. Failing that, you could have
> used a 
> VM that can save ("suspend") the guest state so that you can resume
> later.
> 
> Food for thought for the future, I guess :-)
> 
> 
If he wants to live dangerously he could try sys-process/criu...

Probably would want to spin up another instance of the computation and
test with that and make sure it works correctly first.

LMP


signature.asc
Description: This is a digitally signed message part


[gentoo-user] Re: 2 months into an 8-month computation.

2019-07-11 Thread Nikos Chantziaras

On 11/07/2019 20:59, Alan Grimes wrote:

'ey, I have the 2.3 months into an 8-month computation blues...
[...]
So basically all gentoo updates will have to be done at the end of this 
run, I'm not really sure when, sometime in the December-January timeframe.


I guess you should have written your code in a way that can store 
current state so that it can resume. Failing that, you could have used a 
VM that can save ("suspend") the guest state so that you can resume later.


Food for thought for the future, I guess :-)