[Info-vax] Are queue manager updates written to disk immediately ?

[Johnny Billquist]

On 2013-04-19 16:35, AEF wrote:
> On Apr 18, 11:37 am, Paul Sture <nos... at sture.ch> wrote:
>> In article <kkoss5$ij... at Iltempo.Update.UU.SE>,
>>   Johnny Billquist <b... at softjar.se> wrote:
>
>>> Simon can correct me if I understood things wrong, but this is my
>>> understanding:
>>
>>> Job X ran.
>>> Job X submitted job Y for running tomorrow.
>>> System crash
>>> Job X ran again.
>>> Job X submitted job Z for running tomorrow.
>>> Job X finished
>>
>>> Job Y disappeared. Job Z exists as expected. Job X got two log files
>>> from the two runs. Job X would appear to have completed both times.
>>
>>> My guess is that job X was rerun because it didn't finish, but that is
>>> perhaps not a good enough explanation for the rerun.
>>> There is no obvious explanation at all for job Y not being around.
>>
>>> The "best" explanation might just be that the quemanager work file was
>>> not updated at all, so job X would appear to not have run at all. Thus
>>> the run after the reboot again, and also the loss of job Y.
>>
>> As I understand it, the queue manager file wasn't even updated to
>> reflect that job X had started, let alone finished.  According to the
>> document I quoted up thread, running jobs should be aborted when the
>> machine comes back
>>
>> --
>> Paul Sture
>
> Thanks!
>
> Just one more:
>
> X - entry number not in accounting

Entry number for X have not been given in any of the posts, as far as I 
know. But I'm sure it could be provided, if we really think we need it. 
(Can't say I think it makes any difference if I manage to put an 
explicit value to X.) But I fail to see any connection to any 
accounting. These are queue job numbers.

> Y - entry number 3927
> Z - entry number 3923
>
> Is this right? If so -- fascinating.

Yes...

	Johnny