AWS Thinkbox Discussion Forums

beta10 - RequiredAssets/ScriptDependencies

Deprecated Deadline 6 Bug Reports
#1

Hi,
Submitting job to Deadline with either Required Assets or Script Dependencies via a to:JobInfoFile local file (ie: SMTD submission). I write the file path entries as a comma separated list which is a mixture of absolute UNC paths & drive mapped paths. However, when it gets submitted into Deadline there are 2 issues in the monitor “Required Assets” UI in job properties:

  1. Asset Name = Asset Path! Essentially, as I haven’t manually browsed via monitor for the assets/scripts and instead inputted the information via submission, it has made the asset name the same as the full asset path.

  2. RequiredAssets / Script Dependencies - needs “CheckPathMapping” method applying to it when you are viewing the job properties on another OS monitor. I can’t EDIT the RequiredAssets entries as the file paths are invalid when on another OS.

Finally, if I submit a job with RequiredAssets to Deadline, will it automatically recognise that assets are required and therefore mark the job as “Pending/Assets”? (I’ve noticed that the slave scheduling is already nicely handling NOT picking up the job if 1 or more assets are missing - that bit works fine.)

Mike

#2

  1. This only seems to affect non-Windows Monitors. This should be fixed in the next beta.

  2. What prevents you from editing them? I seem to be able to edit them here just fine. Obviously it won’t default to the file’s location, but you should still be able to edit them.

Yes.

Cheers,

  • Ryan
#3

Hi Ryan,
So, did some more playing with RequiredAssets/Dependencies today.

  1. Cool. Confirmed, display is ok on Win but incorrect on OSX.

  2. Indeed, I can hit the EDIT button and it works. I was thinking maybe we should apply PathMapping to the JobProperties dialog depending on platform it’s being viewed on. If a studio has setup global PathMappings in it’s repo options, then if a user on a different OS than the one the job was submitted on, want to edit a RequiredAsset OR Script Dependency, then I thought it would be useful to display the file paths so that they can easily navigate to these assets? I’m not entirely convinced myself…just explaining my thinking aloud…

  3. If I execute a manual “Perform Pending Job Scan”, the console successfully does any required PathMapping on whatever OS and then releases any pending jobs where all RequiredAssets are now in place. However, if I RC and mark the job as pending again on either Win or OSX and carry out another manual “Perform Pending Job Scan”, the job never get’s released again. Execute the Job Scan as many times as you like, the job isn’t going to get released again.

  4. Once a pending job, based on RequiredAssets is released, if 1 or more assets/script dependencies becomes ‘unavailable’ - let’s say, gets deleted, then the job doesn’t get marked as pending again by the next Job Scan. The Slaves log correctly via scheduler that the job is not de-queued as a particular asset/script is not present, however, looking at just the monitor and assuming I’m remote to all my slaves, how am I going to know that this condition has happened? I was thinking that both the “Pending Job Scan” should re-pend a job if this condition occurs, but also, the monitor needs a visual clue for the user that this is the reason the job isn’t getting picked up. I reckon it’s going to be quite confusing for users to understand/debug why their job isn’t getting picked up! - So, how about, the RequiredAssets/ScriptDependencies UI has a 3rd column which identifies the STATE of each asset/script?

“EXISTS”, “MISSING”, “UNKNOWN” and any missing/unknown are highlighted in RED in the list? OR do we introduce “Pending” job reports to Deadline? ~ seems messy, generating lots of reports every time a Job Scan tales place…?

Also, would be good if the node view also highlighted those assets/scripts which were in trouble.

Thanks,
Mike

#4

  1. I can’t reproduce this. This seems to work just fine here…

  2. Yeah, I’m not sure how to handle this. It might make sense to just remove that additional check from the slaves so that the job just fails due to missing assets, rather than sitting there idling, but that could also be bad if the job doesn’t fail and goes on rendering for hours like everything is okay. Another option is to have the slave re-pend the job itself if it can’t see the assets, but then this can be a problem if one slave can’t see the assets (ie: path isn’t mapped, it’s in a different location, etc), but others can see them just fine.

We do want add a “test” button (or whatever) to the dependency view for Deadline 7 that can be used to test the dependencies.

Cheers,

  • Ryan
#5

  1. WTF! Yeah, I can’t reproduce it now using the same file as before. I’ll keep trying…

  2. Indeed. Tricky. Say, the 3dsMax plugin can be told to IGNORE missing textures, so it would continue to render incorrectly. Yep, drive mappings do drop off, so can’t rely on just what one slave think’s it’s best to mark a job pending again and hence, stop it from processing. We do need to handle this situation tho… :confused:

Privacy | Site terms | Cookie preferences