search

truenas / charts

TrueNAS SCALE Apps Catalogs & Charts
BSD 3-Clause "New" or "Revised" License
290 stars 278 forks source link

Immich - Insufficient gpu.intel.com/i915 #2649

Closed porjo closed 1 month ago

porjo commented 1 month ago

Since updating to 4.0.7 the app is failing to deploy. I can see immich-machinelearning failing with this error

  Warning  FailedScheduling  73s   default-scheduler  0/1 nodes are available: 1 Insufficient gpu.intel.com/i915. preemption: 0/1 nodes are available: 1 No preemption victims found for incoming pod..

The update has added a new config option to specify which containers gets the GPU allocation, and by default both server + machine learning are selected.

I was able to resolve my issue by unselecting one of the containers. Is it possible for both containers to use GPU? (I have allocated all 5 units of GPU to Immich app)

Similar to #2336

stavros-k commented 1 month ago

Can you try allocating 1? This should allocate 1 on each component. Thanks!

porjo commented 1 month ago

Can you try allocating 1? This should allocate 1 on each component.

That worked, thanks! So the GPU resource count is per container, not the whole app. It would be good to clarify that in the UI.

stavros-k commented 1 month ago

On next Scale release which will be running docker, which will be a bit different on how gpu's work! Will update UI on all apps there! Thanks