AWS Thinkbox Discussion Forums

beta 6 feedback

Deprecated Deadline 6 Beta Discussion
#1

  1. new Monitor Layout - awesome! It’s exactly how I would have designed it! Only thing is the layout name field is very short in its width. I would increase the overall width of the dialog OR drop the checkbox UI element to the row below?

  2. It might be an idea to have an option to print out the sys.path, pythonhome & pythonpath env sys variables at the beginning of each job. I’m starting to getting a little fuzzy on what variable is being taken from where and at what stage of the deadline execution process, that these things are happening…

  3. I need to test, but can the API folder in the local Deadline client installer now be removed? Perhaps, it has been already for new installs?

  4. On that note, auto-upgrade on OSX worked between beta 5 & 6. Hooray!

  5. Shotgun looks all sorted now now as well :slight_smile:

  6. azure.py - line:#201 - defaults?

user = self.GetConfigEntryWithDefault( "LoginUser", "deadlineVMX" ) password = self.GetConfigEntryWithDefault( "LoginPassword", "F00bar123" )

  1. I assume more cloud provider plugins will be added shortly? :slight_smile: These would be good (especially vCloud/cloudStack/GCE):

#1. vCloud
a. cloudStack (Apache)
b. Google GCE
c. Terremark (vCloud)
d. Rackspace - (openStack Public)
e. Joyent
f. Savvis / CenturyLink

  1. “The Scan For Missing Output dialog is now disabled for Tile and Maintenance jobs.” - not quite what I asked for…but fair enough. I can see how it’s hard to predict what files will be where and the user could easily set the temp tiles to be deleted after the job, etc…

  2. “Scan for Missing Output” issues;

(a) Shouldn’t the “OK” button really say “Re-queue”…how does a user know that the functionality is there to auto-re-queue based on missing tasks? I mean…I know…as I know Deadline…but other users…?

(b) If I un-check all the tasks OR un-check all tasks AND un-check the job name at the top AND click “OK”, the pop-up still comes up asking to confirm “re-queuing” these tasks!..doesn’t sound right!

  1. I have 2 completed jobs. When I RC each of them individually, one of them has at the top of the RC menu the “requeue job” option and the other doesn’t! Am I missing something here? (Both Max jobs, 1 normal render job, the other a new tile render job)

  2. I have 2 suspended jobs…Same issue as #10 above!

  3. Scripts Menu in “Configure Repository Options…” > “Submission Scripts” - If I click on a particular script/name and click on the script icon button (the icon has already been assigned to an ico file), if I click cancel on the file browser dialog, it deletes the previous assigned icon to this particular script. Also, please can the file browser dialog remember either the last directory navigated to…man…it nearly gave me RSI, browsing for every single icon file for every single script! AND when you click on the pre-existing icon image/button, can it open the file browser to that current location of the assigned icon file? - Would make life much easier!

  4. “Added ‘Find’ (CTRL-F) functionality to the report panels, history dialogs, and the console panel.” - on OSX, the keyboard shortcut isn’t working (cmd + f) for report panel or console panel or remote commands. However, it is working in the history dialog? The RC menu is there well. Also the form title is: “Form”…probably should be “Search For…” or something!

  5. VRay Spawner executables don’t have some default, Deadline shipping exe file paths for Rhino or SketchUp. Normally you guys do this…

  6. “Added functionality to view/export submission parameters (ie, Job/Plugin info file contents) for a job, in the Job Properties dialog.” - this is awesome! working really well. debugging for a TD just got a lot easier!

  7. SMTD - tiles rendering - “compose over previous output” is still enabled by default. I thought this was going to be un-checked by default?

  8. “Connect to Slave Log…” - RC option on a rendering slave…it may just be me, but is this broken? Going to the actual slave and clicking on slave log button and all is well.

  9. Would be good if by default, when you open a job reports panel, that the tasks were by default “sorted” by the first task (task: 0) at the top.

  10. Draft Tile Assembler - STDout improvement:

2013-09-26 15:16:30: 0: STDOUT: Draft 1.0.1.50572 2013-09-26 15:16:30: 0: STDOUT: Error on missing tiles = True 2013-09-26 15:16:30: 0: STDOUT: Error on missing Background = True 2013-09-26 15:16:30: 0: STDOUT: Distance as pixels = True 2013-09-26 15:16:30: 0: STDOUT: Assembling Single File 2013-09-26 15:16:30: 0: STDOUT: Final Image Dimensions = 640X480 2013-09-26 15:16:30: 0: STDOUT: TileCount given: 34 2013-09-26 15:16:30: 0: STDOUT: Tile Tile0 position: 0.0, 0.75 2013-09-26 15:16:30: 0: STDOUT: Compositing tile Tile0 2013-09-26 15:16:30: 0: PYTHON: 0.03 2013-09-26 15:16:30: 0: STDOUT: Progress: 3 2013-09-26 15:16:30: 0: STDOUT: Tile Tile1 position: 0.25, 0.75 2013-09-26 15:16:30: 0: STDOUT: Compositing tile Tile1 2013-09-26 15:16:30: 0: PYTHON: 0.06 2013-09-26 15:16:30: 0: STDOUT: Progress: 6 2013-09-26 15:16:30: 0: STDOUT: Tile Tile2 position: 0.5, 0.75 2013-09-26 15:16:30: 0: STDOUT: Compositing tile Tile2 2013-09-26 15:16:30: 0: PYTHON: 0.09 2013-09-26 15:16:30: 0: STDOUT: Progress: 9 2013-09-26 15:16:30: 0: STDOUT: Tile Tile3 position: 0.75, 0.75 2013-09-26 15:16:30: 0: STDOUT: Compositing tile Tile3 2013-09-26 15:16:30: 0: PYTHON: 0.12 2013-09-26 15:16:30: 0: STDOUT: Progress: 12

Cool! Maybe tweak progress to print - “Progress: 12%” and update Deadline’s “Task Render Status” field with this information saying something like: “Draft Assembling Single File - 12%”? For larger images, it will be good to know exactly what each slave is currently doing. Also, “Final Image Dimensions = 640X480” could just be “Final Image Dimensions = 640x480” to display the resolution. Sorry! picky, it just sticks out when I look at it!

  1. see attached image. I’ve had this a few times now. Hard to reproduce. Best guess is its when I try to RC the windows deadline launcher too fast and click on slave to start-up and communication with MongoDB for some reason has a little fit / network blip or can’t communicate with MongoDB and then errors as per the image. Dismissing the dialog and trying to start MongoDB up again and all is well…if I go to a web browser and mbp.local:28017, all is well…

Untitled.png
Untitled.png735×511 301 KB

  1. “Scan For Missing Output” may not be visible for 3dsMax tile rendering jobs now, but it is still visible for Draft Tile Rendering jobs…

  2. My Draft Tile Rendering jobs are now working correctly :slight_smile:

#2

Thanks for the feedback!

  1. We’ll make the initial width of the dialog wider.

  2. Here’s how it works now. When a slave starts a new process, it passes the original system PYTHONHOME and PYTHONPATH environment variables to the rendering process, so that it’s basically the same environment as when you’re running the process outside of Deadline. The exception to this is if a PYTHONHOME or PYTHONPATH variable is set in the job’s environment variables. In this case, the job’s variables take precedent.

  3. Yeah, it’s already been removed for new installs. You can remove it on your end if you want.

  4. Yay! :slight_smile:

  5. See (4).

  6. Oops. The defaults should be empty (""). We’ll get that fixed.

  7. I’ll pass this on to the VMX guys. I’m sure we’ll add more in the future, but we want to keep the initial scope small.

  8. Yeah, with us moving to the Draft config file system, Deadline itself won’t be aware of the output paths of each individual tile.

  9. a) We’ll add some info at the top of the dialog explaining how it works, as it’s not always clear you need to press “Scan” either.
    b) This has already been fixed internally, and will be included in beta 7.

10/11. Can you export the jobs that aren’t working properly and post them? We can try importing them to see if we can reproduce.

  1. Yeah, we need to make some usability improvements here. We’ve logged this as a bug. We need a better way to clear the current shortcut too.

  2. Logged as a bug.

  3. We don’t actually know the defaults at this time, so we’re leaving them blank for now. :slight_smile:

  4. See (4).

  5. Hmm, I thought we had done this, but I guess not. We’ll get that fixed!

  6. It’s working here for me. Is the panel just not popping up? Or does it popup but fails to connect?

  7. Currently, it sorts by data/time so that the most recent report is at the top. This is one of those cases where we won’t be able to please everyone with a good default. If we can improve how defaults are applied to panels when they are opened like we had discussed a little while ago, this will probably be moot (viewtopic.php?f=156&t=10181).

  8. I’ll pass this feedback to Grant. Thanks!

  9. We’ve seen this before too. Definitely need a “cleaner” error in this case!

  10. Yeah, that’s because we’re currently looking at the job mode (if it’s Normal, Tile, or Maintenance). The assembly jobs are still Normal jobs. We want to keep the file info with the job so that you can still explore the output or open an image viewer for this job. Might just need to be one of those edge cases we need to live with for now…

  11. See (4).

Cheers,

  • Ryan
#3

  1. OK, understood.

  2. Really? I was not thinking this!

  3. See attached. Unfortunately, testing the auto-delete job function, the suspended job example was deleted and purged by Pulse, so I only have 1 example at present for you. Hopefully, this will be enough. If not, give me a shout. (Had to remove the actual Max file as it contains data that doesn’t belong to me)

owenm__3dsmax__L3Q_INT_MP0070_PRVZ_A01.zip (6.33 KB)
17. Ignore me. Networking issue. The panel pops up fine.

  1. Yeah, understood, but I guess you see where I’m coming from as well. With the new monitor UI layout global system in place, how about adding the column sorts custom configuration to the layout system. It get’s loaded at startup or reset at startup and that’s it for that monitor session for that user. If the user changes something, it’s up to them, but will potentially get reset on the next monitor session? OR how about only allowing 1 panel of each type to be visible at 1 time. ie: multiple “jobs” panels with different filtering, but they all have the same default column sort order to start off with when opened for that particular monitor session? Maybe, I’m over-complicating the whole situation here?! :confused:

  2. So, I don’t really understand the error. The slave (machine) is unable to connect to the MongoDB? I find this quite hard to believe as this is on a windows VM and the DB is running on the same physical machine (OSX), so no ‘real’ network slowdown or ‘real’ network blip…my previous suggested reason for the error…was nothing more than an educated guess…perhaps when slave tries to connect to the DB, let it have another chance after a couple of seconds and then exit out if it still can’t connect? (I’m like 99% sure the MongoDB is running whenever I start the windows based slave, as I will already have at least OSX based monitor up and visible and if that can’t connect to the DB, then I would see it in the monitor). - Perhaps, this error has more to do with networking connectivity ‘specifically’ when it comes to VM’s which might need investigating further? - I should note that the Slave was NOT on auto-startup when launcher starts…so it’s not like Slave tried to start to soon as the OS booted up and networking wasn’t quite ready…

#4

  1. Thanks! We’ll take a look.

  2. Yeah, I’m sure there is a elegant solution to this issue. I’m sure we’ll think of something, but it may have to wait until 6.2.

  3. It’s hard to see in all the mess, but the error is saying it can’t resolve the host name for the machine that mongo is running on (mbp.local). Maybe a temporary DNS issue? At the very least, this error message will be much cleaner in beta 7. :slight_smile:

#5

I’m pretty sure those were set intentionally so that it would still work even if you didn’t configure a login. Azure complains if you don’t set one – actually, now that I think of it, there should probably be a symbol in there to satisfy their password requirements.

I guess I could see the argument for just forcing users to specify a password, though… If that’s the case we should definitely just print out an error ourselves ahead of actually making the Azure call, I guess.

Privacy | Site terms | Cookie preferences