Hi there
I have been trying to access a 2 GPU instance for almost 10 days now but I keep getting the following error:
You session cannot be scheduled due to insufficent resources. The most likely reason is: ‘Insufficient nvidia.com/gpu’. You may wait for resources to free up or you can adjust the specific resource and restart your session.
Could you please confirm that the session will automatically launch once the resources will be available and that I don’t have to keep creating new “requests”?
Also, can you estimate how much longer all resources will be blocked?
I am approaching a project deadline and I need to run some trials that will likely take ~3 days using 2 GPUs.
Thank you,
Raphaela
I’m guessing you are using limited, the ETH / EPFL internal deployment of Renku. That instance only has 6 GPUs in total, making them very scarce; the session you want to launch is requesting 1/3 of the potentially available GPUs, although it will indeed start when the resources become available, it could be a while before that happens.
If you want to reach out, either via DM or through a contact you may have at SDSC, we can see if there is something that can be done to assist in this situation.
Hi @rwagner, I double checked the deployment you are using from our end. We had recently added some gpus so we currently have 8 gpus available there. All are currently taken.
If you leave your session in the state as what you showed in the screenshot above, when gpus become available it will get scheduled. However I cannot tell when this will happen. 8 gpus is really not a lot so they are utilized fully almost all of the time. When your session will be scheduled depends on when will other users close their sessions that use gpus and I cannot predict this.
Okay thank you both for your responses. I’ll let my supervisor know and will keep waiting.
Best, Raphaela
