It’s possible to change the region/zone when the instance is in status Terminated ?
Previously I was working with gradient solution from Paperspace but my notebooks were deleted after each restart, now after change to GCP platform I can’t start the instance so I’m wondering if I should try with a new platform or create a new instance in a different region/zone. AFAIK in both cases I will start from scratch again
2 Likes
Trying a lot from the last couple of hours but got the same error. Already on upgraded account
Starting VM instance “my-fastai-instance” failed. Error: The zone ‘projects/driven-rider-220513/zones/us-west2-b’ does not have enough resources available to fulfill the request. Try a different zone, or try again later
2 Likes
is this due the preempitble type of VM
I created another instance frm scratch with different instance name and zone us-west2-c and it worked
I am getting following error
ssh: connect to host 35.235.85.224 port 22: Connection refused
ERROR: (gcloud.compute.ssh) [/usr/bin/ssh] exited with return code [255]
I have tried:
- Creating new ssh key and adding on the server
- Using the ‘upgrade’ option
- Creating a new instance on a different Gcloud account
- Removing --preemptible
- Changing zones (to overcome Try a different zone, or try again later error)
Any suggestion to fix this is welcome, thank you.
Maybe, there is a firewall blocking, have you checked that? Try using your phone 4g/3g to test it!
Me either…
I already created 2 instances in different zones, and both don’t start.
Also upgraded my account.
There is no way to make them work!
I opened a ticket at google support and they said they would monitor bla bla bla…
NOTHING
I wish someone from google could monitor this thread and help us…
3 Likes
If you haven’t already then try different zones which support your GPU:
https://cloud.google.com/compute/docs/gpus/
1 Like
Use us-west2-c with the p4 GPU for now.
That’s the only zone which has resources.
We all know currently we are having some issues related to availability of resources.
So, I just add a new utility on fastai-shell to test that.
Based on my testing with that us-west2-c is the only zone which has resources. That might change.
You can try it yourself by running my utility.
Hope this helps.
3 Likes
Thanks Arunoda. I was able to create a new instance at us-west2-c like you said.
Let´s see how long this one will work correctly…
Is the issue with available resources still an issue with preemptibility disabled?
Anyone tested?
Yes, the issue persists with preemptibility disabled.
1 Like
I checked the firewall setting on the server and looks it should accept the connection
default-allow-ssh Ingress Apply to all IP ranges: 0.0.0.0/0. tcp:22 Allow 65534 default
But when I checked with nmap from my local machine I get,
PORT STATE SERVICE
22/tcp closed ssh
The instance is running.
how do you disable the preemptible option?
When you use the instructions in fastai for creating a new instance, simply skip the --preemptible option in step 3. https://course-v3.fast.ai/start_gcp.html
If the instance is already created is there a way to change it?
3 Likes
Even when you seem to get the instance going somehow I feel the compute is slow.Is anyone else experiencing that?
Could anyone get the lesson3 NLP notebook to run with the us-west2-c instance? I lowered the batchsize to 30 but still got cuda out of memory errors more than half an hour into the training.