Unable to run Camunda Modeler on Fedora 35 VM

BPMN & Modeling Discussion & Questions
1

Hi,

I am trying to run Camunda Modeler in a Fedora 35 VM.

$ curl -SLO https://downloads.camunda.cloud/release/camunda-modeler/4.12.0/camunda-modeler-4.12.0-linux-x64.tar.gz

$ cd camunda-modeler-4.12.0-linux-x64
$ ./camunda-modeler
INFO app:main:bootstrap starting Camunda Modeler v4.12.0
INFO app:cli parsing [ '/home/dan/camunda-modeler-4.12.0-linux-x64/camunda-modeler' ] in '/home/dan/camunda-modeler-4.12.0-linux-x64'
INFO app:flags searching for flags.json in paths [
  '/home/dan/camunda-modeler-4.12.0-linux-x64/resources',
  '/home/dan/.config/camunda-modeler/resources'
]
INFO app:flags found []
INFO app:flags active {}
INFO app:error-tracking Cannot initialize Sentry:  Crash reports not enabled.
INFO app:plugins searching for plugins/*/index.js in paths [
  '/home/dan/camunda-modeler-4.12.0-linux-x64',
  '/home/dan/camunda-modeler-4.12.0-linux-x64/resources',
  '/home/dan/.config/camunda-modeler/resources',
  '/home/dan/.config/camunda-modeler'
]
INFO app:plugins found plug-in entries []
INFO app:plugins registered []
The futex facility returned an unexpected error code.
Aborted (core dumped)

I read another post in this forum where a user was experiencing a similar issue - setting DISPLAY=:0 was recommended. This seems to get me a bit further:

$ export DISPLAY=:0
$ ./camunda-modeler 
INFO app:main:bootstrap starting Camunda Modeler v4.12.0
INFO app:cli parsing [ '/home/dan/camunda-modeler-4.12.0-linux-x64/camunda-modeler' ] in '/home/dan/camunda-modeler-4.12.0-linux-x64'
INFO app:flags searching for flags.json in paths [
  '/home/dan/camunda-modeler-4.12.0-linux-x64/resources',
  '/home/dan/.config/camunda-modeler/resources'
]
INFO app:flags found []
INFO app:flags active {}
INFO app:error-tracking Cannot initialize Sentry:  Crash reports not enabled.
INFO app:plugins searching for plugins/*/index.js in paths [
  '/home/dan/camunda-modeler-4.12.0-linux-x64',
  '/home/dan/camunda-modeler-4.12.0-linux-x64/resources',
  '/home/dan/.config/camunda-modeler/resources',
  '/home/dan/.config/camunda-modeler'
]
INFO app:plugins found plug-in entries []
INFO app:plugins registered []
INFO app:main:bootstrap received ready
libva error: vaGetDriverNameByIndex() failed with unknown libva error, driver_name = (null)
libva error: vaGetDriverNameByIndex() failed with unknown libva error, driver_name = (null)
libva error: vaGetDriverNameByIndex() failed with unknown libva error, driver_name = (null)
libva error: vaGetDriverNameByIndex() failed with unknown libva error, driver_name = (null)
libva error: vaGetDriverNameByIndex() failed with unknown libva error, driver_name = (null)
libva error: vaGetDriverNameByIndex() failed with unknown libva error, driver_name = (null)
[5332:0221/151751.203749:FATAL:gpu_data_manager_impl_private.cc(415)] GPU process isn't usable. Goodbye.
Trace/breakpoint trap (core dumped)

EDIT: The same process works fine on Ubuntu 20.04. Does anyone know why I’m seeing this error with Fedora?

2

So I found a fix for this in the end. This seems to be related to the version of electron that is being used and an update to glibc that caused issues with the seccomp based chrome sandbox that is used by electron.

To run Camunda Modeler 4.12.0 on Fedora 35 you need to add a flag to get around the issue above:

./camunda-modeler --disable-gpu-sandbox

Hopefully, an update to the version of electron used in the Camunda Modeler package will fix this issue for future releases.