From: "Abraham Schneider" <aschneider@(email surpressed)>
Subject: Bug with jobdonecommand/dump?
   Date: Mon, 23 Jun 2008 07:29:49 -0400
Msg# 1758
View Complete Thread (6 articles) | All Threads
Last Next
Hi!

We updated to the new version last week. FIFO seems to work very fine and helps alot for our situation here! Thanks for that!

But today I recognized a strange behaviour with rush. I'm not sure if this is a 'bug/feature' of the new version or if this is an old behaviour. When I dump a job which hasn't finished yet, there is a moment where there is the text in the remarks of this job that it executes the jobdonecommand.

Is this an intentional behaviour or a bug? For me, it makes no sense to execute a script/command in this situation which is meant to run when the job has finished rendering completely without any problems. On the contrary, I could think of situations where this behaviour could lead to problems.

Best regards,

Abraham

--
Abraham Schneider
Senior VFX Compositor
ARRI Film & TV Services GmbH  Tuerkenstr. 89  D-80799 Muenchen / Germany
Phone (Tel# suppressed) Email aschneider@(email surpressed)
www.arricommercial.de
www.arri.de

ARRI Film & TV Services GmbH
Sitz: München  -  Registergericht: Amtsgericht München  -  Handelsregisternummer: HRB 69396
Geschäftsführer: Franz Kraus

   From: Greg Ercolano <erco@(email surpressed)>
Subject: Re: Bug with jobdonecommand/dump?
   Date: Wed, 25 Jun 2008 00:55:49 -0400
Msg# 1765
View Complete Thread (6 articles) | All Threads
Last Next
Abraham Schneider wrote:
Hi!

We updated to the new version last week. FIFO seems to work very fine=20
and helps alot for our situation here! Thanks for that!

But today I recognized a strange behaviour with rush. I'm not sure if=20
this is a 'bug/feature' of the new version or if this is an old=20
behaviour. When I dump a job which hasn't finished yet, there is a=20
moment where there is the text in the remarks of this job that it=20
executes the jobdonecommand.

    Thanks Abraham -- I'll have to check into this to see if this
    is intended behavior or not, and if not, at what point it crept in,
    and fix it.

    When I find out more, I may come out with a follow up fix release
    to solve it.

    I can confirm 'jobdonecommand' is indeed being executed when a job
    is dumped even if the frames have not finished.

    For the short term, your jobdonecommand can look at the 'framelist'
    file to determine if there were unfinished frames or not.

    When I return next week, I'll definitely look into this.

    One question: what version of Rush did you upgrade /from/?

Is this an intentional behaviour or a bug? For me, it makes no sense to=20
execute a script/command in this situation which is meant to run when=20
the job has finished rendering completely without any problems.

    Actually, I need to check to see if the docs ever said that the
    jobdonecommand command would not be run if there were any 'Fail'
    frames -- I'm not sure that's the case.

    I'd agree certainly the jobdonecommand shouldn't be executed there
    are still frames in the "Que" or "Hold" state.

    But whether any frames were 'Fail'ed or not, the jobdonecommand
    can look at the framelist or 'rush -lf/-lfi' report to make that
    determination.

    There may be cases where it's desirable for the jobdonecommand
    to be executed even if some frames Fail(ed), such as generating
    a Quicktime with the failed frames replaced with "X" frames
    to give the TDs a hint as to which frames failed, and to go back
    and check the logs for those frames to see what went wrong.

    What I should maybe do is add a flag for the jobdonecommand
    that defines whether rush should execute the command only
    if all the frames are 'Done', or 'Done/Fail'.

--
Greg Ercolano, erco@(email surpressed)
Rush Render Queue, http://seriss.com/rush/
Tel: (Tel# suppressed)
Fax: (Tel# suppressed)
Cel: (Tel# suppressed)

     ********************  N O T I C E  *************************
     ***   I'll be on vacation June 18th through June 28th.   ***
     ***   During this time I will be watching email, but it  ***
     ***   may take up to a day for me to reply.              ***
     ************************************************************

   From: "Abraham Schneider" <aschneider@(email surpressed)>
Subject: Re: Bug with jobdonecommand/dump?
   Date: Wed, 25 Jun 2008 10:03:08 -0400
Msg# 1766
View Complete Thread (6 articles) | All Threads
Last Next
Greg Ercolano schrieb:
 >     One question: what version of Rush did you upgrade /from/?

It was 102.42a6, I think.

    What I should maybe do is add a flag for the jobdonecommand
    that defines whether rush should execute the command only
    if all the frames are 'Done', or 'Done/Fail'.


That would be a good solution, I think.

Thanks, Abraham


--
Abraham Schneider
Senior VFX Compositor
ARRI Film & TV Services GmbH  Tuerkenstr. 89  D-80799 Muenchen / Germany
Phone (Tel# suppressed) Email aschneider@(email surpressed)
www.arricommercial.de
www.arri.de

ARRI Film & TV Services GmbH
Sitz: München  -  Registergericht: Amtsgericht München  -  Handelsregisternummer: HRB 69396
Geschäftsführer: Franz Kraus

   From: Greg Ercolano <erco@(email surpressed)>
Subject: Re: Bug with jobdonecommand/dump?
   Date: Mon, 07 Jul 2008 12:17:16 -0400
Msg# 1772
View Complete Thread (6 articles) | All Threads
Last Next
Abraham Schneider wrote:
>>     What I should maybe do is add a flag for the jobdonecommand
>>     that defines whether rush should execute the command only
>>     if all the frames are 'Done', or 'Done/Fail'.
> 
> That would be a good solution, I think.

	Done; I've added this to 102.42a10.

	I've currently implemented it as optional flags to the jobdonecommand,
	similar to the current '-nolog' flag, eg:

		-done     -- run only if all frames are 'Done'
		-donefail -- run only if all frames are either 'Done or Fail'
		-fail     -- run only if /any/ frames are 'Fail'
		-always   -- run the command under any condition, even if frames 
		             aren't finished (ie. when job dumped mid-render)

	Abraham, at some point I can supply you with binaries for this new
	version after I run some tests.

-- 
Greg Ercolano, erco@(email surpressed)
Seriss Corporation
Rush Render Queue, http://seriss.com/rush/
Tel: (Tel# suppressed)
Fax: (Tel# suppressed)
Cel: (Tel# suppressed)

   From: Andrew Kingston <andrew@peerless.co.uk>
Subject: Re: Bug with jobdonecommand/dump?
   Date: Thu, 09 Dec 2010 07:01:42 -0800
Msg# 1975
View Complete Thread (6 articles) | All Threads
Last Next
On 07/07/2008 17:17, Greg Ercolano wrote:
Abraham Schneider wrote:
     What I should maybe do is add a flag for the jobdonecommand
     that defines whether rush should execute the command only
     if all the frames are 'Done', or 'Done/Fail'.

That would be a good solution, I think.

	Done; I've added this to 102.42a10.

	I've currently implemented it as optional flags to the jobdonecommand,
	similar to the current '-nolog' flag, eg:

		-done     -- run only if all frames are 'Done'
		-donefail -- run only if all frames are either 'Done or Fail'
		-fail     -- run only if /any/ frames are 'Fail'
		-always   -- run the command under any condition, even if frames
		             aren't finished (ie. when job dumped mid-render)

	Abraham, at some point I can supply you with binaries for this new
	version after I run some tests.

Hi

I'm running 102.42a9, but I really need the "-done" option on my job servers. Is it possible to just upgrade certain machines to 102.42a10?

Cheers
Andrew

   From: Greg Ercolano <erco@(email surpressed)>
Subject: Re: Bug with jobdonecommand/dump?
   Date: Thu, 09 Dec 2010 15:55:57 -0500
Msg# 1976
View Complete Thread (6 articles) | All Threads
Last Next
Andrew Kingston wrote:
>>> That would be a good solution, I think.
>> 	Done; I've added this to 102.42a10.
>>
>> 	I've currently implemented it as optional flags to the jobdonecommand,
>> 	similar to the current '-nolog' flag, eg:
>>
>> 		-done     -- run only if all frames are 'Done'
>> 		-donefail -- run only if all frames are either 'Done or Fail'
>> 		-fail     -- run only if /any/ frames are 'Fail'
>> 		-always   -- run the command under any condition, even if frames
>> 		             aren't finished (ie. when job dumped mid-render)
>>
>> 	Abraham, at some point I can supply you with binaries for this new
>> 	version after I run some tests.
>>
> I'm running 102.42a9, but I really need the "-done" option on my job 
> servers.  Is it possible to just upgrade certain machines to 102.42a10?

	a10 hasn't been released yet.

	Note that you can get this already if you have your current
	donecommand script check the status of the job to see if it's
	Done or has Fail frames.

	When your donecommand runs, it will have various environment variables
	set telling you e.g. what the jobid is ($RUSH_JOBID).

	You can, for instance, use 'rush -lfi <JOBID>', or 'rush -lj <JOBID>',
	or you can look at the 'framelist' or 'jobinfo' files in the log directory
	to determine if the job completed with Done or Fail.


-- 
Greg Ercolano, erco@(email surpressed)
Seriss Corporation
Rush Render Queue, http://seriss.com/rush/
Tel: (Tel# suppressed)ext.23
Fax: (Tel# suppressed)
Cel: (Tel# suppressed)