What happened to my datacenter?

[half hemlock]

Hi everyone, I'm new here, I only discovered Runpod yesterday.
I created a Network Volume on the EUR-NO-1 datacenter and yesterday the availability of RTX 5090s was shown as "HIGH". Today, when I was supposed to start working, all the GPUs in the datacenter are "Unavailable". Is this normal? What does "Unavailable" actually mean, have the GPUs been removed from the datacenter, or are they all being used by other users?
Is this something that happens often, or have I just been unlucky?

It would be helpful if the website clearly communicated GPU status and acknowledged when issues are occurring.
I'm currently spending credits on the network volume without being able to create any Pods on it, which is quite frustrating.

[alpine scarab]

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

[split cipher]

I see lot of 5090 on that region. Do you use some filters?

[half hemlock]

Hey thank you for answering me.
I don't think I am using any filter, that is what I see:

I tried switching to NVME from SSD but it's the same

[whole cedar]

I'm confirming with the team but my impression is that the availability indicators can be a bit misleading on certain high volume GPU specs like the 5090 - because it appears to be a flat amount that triggers it as 'high availability' but there's a difference between there being 10 available on a spec where we have dozens total, and 10 available on a spec where we have thousands. The latter is much more prone to go from 'high' to unavailable.

I tried to start in the same DC myself and it is showing as unavailable for me too so it looks like they are just maxxed out. I will bring the feedback to the team to hopefully be a little more in line with customer expectations on what 'high availability' means

[half hemlock]

ok I understood thank you for taking the time to answer me.

Today the GPUs are back on my datacenter. I hope they update the drivers though

currently CUDA 13 is not supported on the GPUs of this datacenter and I already downloaded all the models on this network volume

[granite kiln]

Yup, mi300x is always low even if every single one is available for the same reason

[whole cedar]

You can use runpodctl to move everything to another network volume if you need - https://www.youtube.com/watch?v=gnSLRrlBfcA