AWS Thinkbox Discussion Forums

Nuke almost random missing frames on first frame of write node in a batch

Hey everyone,
We have this strange random error occurring. I cannot find a pattern, but when rendering through Nuke there are random frames that don’t get written. I’d say 80% have no issues but then I get these random errors of…
RuntimeError: Write2: Unable to create directory
The first frame of the batch will get skipped and then all the other frames in the batch are fine. I don’t see a pattern on why this is an issue. Anyone have any suggestions? Again this isn’t consistent and maybe every 5th render sees these symptoms. Thank you in advance for your help.

Right now I can’t tell if this is a Deadline issue or a Synology issue. It’s so random and I cannot repeat it.

Hello @chris_m_wood

I think we can run a command line job in a loop to test the storage.

If loading the scene/file/asset breaks randomly and you suspect it’s due to the performance issue of the storage device run below tests to prove/disprove yourself:

Test1:

  • Submit a command line job: Deadline Monitor> Submit> Misc.> Command Line
  • Executable is: /bin/dd and the Arguments: if=/path/to/scene of=/dev/null bs=1024000
  • frame list 0 - 100 (make enough Tasks to so that every machine can render once )
  • This job will be pulling files from the disks (if) of size (bs) 1024000 bytes and prints the I/O rate, it will copy the files to a null location (of)

Test2:

  • Do like the Test:1 but increase the number of concurrent tasks job and Worker. In this test we would know if more copy operations per machine is causing the performing issues.
Privacy | Site terms | Cookie preferences