Re: New behavior proposal --halt -% with job killing

[Martin d'Anjou]

On 15-04-25 03:25 AM, Ole Tange wrote:
> On Thu, Apr 23, 2015 at 5:10 PM, Martin d'Anjou
> <martin.danjou14@gmail.com> wrote:
>
> Good summary:
>
> > You could come up with many options here:
> >
> > When to halt:
> > --halt [condition for halt]
> > --timeout [condition for halt is an amount of time]
> > --memfree [condition for halt is an amount of memory]
> > kill -TERM [condition for halt is the signal]
> I think our solution should make it possible to extend this list. E.g.
> maybe it will be possible to detect whether the remote job failed or
> the network connection to the remote server failed.
>
> --memfree is special, however. It retries indefinitely, if the job
> gets killed due to low memory.
>
> > How to handle jobs after a halt:
> > --halt-job-handling [killpending[,killrunning]]
> > --timeout-job-handling [killpending[,killrunning]]
> > and so on. Users could use both kills if they wanted both.
> And --kill-TERM-job-handling
>
> killrunning will always imply killpending, but the opposite is not the
> case, right?

For me yes, killrunning would always imply killpending. So it could be 
[killpending,killall].

> --retries should be thrown into the mix, too.

What does retry mean?

> I can easily think of real life situations where the handling of a
> death due to --memfree is different from a --timeout, and I think the
> current behaviour (retrying indefinitely) is correct. But do we have a
> real life situation where we want --halt-job-handling to be different
> from --timeout-job-handling given that we have --retries?
>
> I am reluctant to put in 3 options that are extremely rarely used
> (there are plenty of options as it is and testing becomes harder the
> more combinations needs to be tested).

Agreed.

For me, one option suffices for all possible ways of halt/timeout/kill 
-TERM: it is either to kill pending (stop spawning new jobs), or kill 
all (which implies kill pending). I do not use memfree so throw it in 
the mix where it makes sense.

> > Or, you could use an explicit "plus" sign to mean halt and kill all running
> > and pending jobs:
> > --halt +1-99%
> POLA would say that --halt +1-99% == --halt 1-99%

There is a precedent. The find program uses +n, -n and n (man find). 
"find /path/to/files* -mtime +5" is not the same as 
"find /path/to/files* -mtime -5".

Martin