AWS AMI available for testing

I tried and there are some unmerged files. I will work on this and get back to you sometime on Friday (US date).

Worked great for me!

Mild hiccup: environment.yml in the repo was modified, so I had to

cd fastai
git checkout -- environment.yml
git pull

to update the repo.

Maybe this is affecting @bdekoven?

2 Likes

Can we follow instructions from V1 to try it?

Works well. Didn’t have to git pull. Thanks!

yes @claytonjy , I had this issue with environment.yml

After fixing “git pull” indicated already up to date. But still have runtime error

What kind of AWS instance did you choose? I chose p2.xlarge; I think any p2.* should work. Did you choose a non-GPU instance? Could explain the error coming from CUDA.

Yeah good point - anything other than p2 won’t work.

I chose t2.micro without paying attention to the AMI description.

I had to put in a request to AWS for access to p2.xlarge.

Thank you @claytonjy for these suggestions!

2 Likes

is it based on geo-location or anyone can access it?
Like if i set my region to US West?

You need to spin up your machine in US West Oregon Region to use the AMI. Anyone can run a machine in any EC2 regions. But you need to remember where you spin up your instances, so you can stop it when not using it. EC2 dashboard has no “Global” View. You have to choose the view from the dropdown in the top right. So be careful and I would suggest you just stick to US West Oregon for all instances. US Regions typically have the lowest cost in AWS - https://aws.amazon.com/ec2/pricing/on-demand/

1 Like

Hello All…
I launched the instance and ssh’ed into it through putty successfully . But I am not able to open jupyter notebook in my browser. It’s running into the instance but when I am trying to open it into my browser it doesn’t connect…

have a look at the discussion ahead in this thread…

Already tried all that… still not being able to open it.
Displays ERR_TIMED_OUT The site can’t be reached

i am also having the same problem like them but i am using root user to do ssh…

root@DESKTOP-V3IQLHT:~/.ssh# ssh -i .ssh/aws_key_pair.pem ubuntu@ec2-35.161.145.248 -L 8888:127.0.0.1:8888
ssh: Could not resolve hostname ~35.161.145.248: Name or service not known

root@DESKTOP-V3IQLHT:~# ls -a
.  ..  anaconda3  .bash_history  .bashrc  .bashrc-anaconda3.bak  downloads  id_rsa_1  .local  .nano  .profile  .ssh
root@DESKTOP-V3IQLHT:~# clear
root@DESKTOP-V3IQLHT:~# ls -a
.  ..  anaconda3  .bash_history  .bashrc  .bashrc-anaconda3.bak  downloads  id_rsa_1  .local  .nano  .profile  .ssh
root@DESKTOP-V3IQLHT:~# cd .ssh/
root@DESKTOP-V3IQLHT:~/.ssh# ls
aws_key_pair.pem  id_rsa_1.pub

What Should i do?
@ramesh

Update

ssh is successful with this one though

root@DESKTOP-V3IQLHT:~# ssh -i .ssh/aws_key_pair.pem ubuntu@ec2-54-70-111-75.us-west-2.compute.amazonaws.com
1 Like

Just spin up the instance p2.xlarge at US:west Oregon. SSH into the instance, ran Git Pull, ran Notebook 1. And got the results just fine :slight_smile:

Thanks Jeremy for this awesome AMI !

6 Likes

I think your issue originally was not using the right IP address; ec2-35.161.145.248 is neither a domain name nor an IP address, and I suspect 35.161.145.248 is your private IPV4 address, not the public one. Also, if you’re already in the .ssh folder, you don’t want the path to the private key to start with .ssh/. I usually give absolute paths (~/.ssh/aws_key_pair.pem) so I can re-run commands from history without caring what my pwd is.

Now that you’ve connected over SSH, you can see that tunnelling the jupyter-notebook port isn’t necessary to connect. If you do want to tunnel that connection over ssh, you can try the working ssh command again, but with -L 8888:127.0.0.1:8888 appended like you tried originally.

2 Likes

Actually I just removed the ec2 and it’s working…

Getting this error while launching jupyter…

channel 3: open failed: connect failed: Connection refused

@ramesh

Basically the error means that no Application in your EC2 instance is listening on port 8888. Did you start Jupyter notebook inside the Remote machine?

Try the command jupyter notebok list, if there’s jupyter server running in EC2, you will see it, otherwise it will come up empty. If it’s empty run Jupyter Notebook.

Because you are using Tunneling of ports, if for any reason you loose the SSH connection, your notebook is not accessible. It might be better to open up port 8888 on the server by modifying the Security Groups (on EC2) and then directly connect via http://<ec2-servername>:8888

2 Likes

Thanks for the info…

wrong thread