Platform: Google Cloud Platform (GCP)

steef · March 24, 2020, 12:55am

Thanks! I deleted my instance and started over with a clear instance without --preemptiple which seems to work so far…

steef · March 24, 2020, 12:56am

Yes! Actually your instructions above (earlier in this thread) are amazing; I followed those and everything is looking great so far Thanks!

vamsiuppala · March 28, 2020, 10:52pm

I am unable to open this link to access the notebooks:
http://localhost:8080/tree.
I can see the instance running from the console. And am able to access them through the terminal too. Not sure how to fix this.
When I try to connect using this command -
gcloud compute ssh --zone=“us-west1-b” jupyter@“my-fastai-instance” – -L 8080:localhost:8080,
it connects but with an message that says “could not request local forwarding”

Can someone please help?

vamsiuppala · March 28, 2020, 11:08pm

I dug around a little to find what local forwarding is.

Landed here: https://manas.tungare.name/blog/ssh-port-forwarding-on-mac-os-x
Used the ‘lsof -i -P’ command to find existing ‘jobs’ that had a ‘8080’ description (I really don’t know if any of this terminology is accurate).
Used the 'kill -9 ’ to kill those jobs
Reconnected using the 8080 port.
And it works now.
Sometimes, I feel like most of what I do is just hoping that something works out.

steef · March 30, 2020, 12:45am

@jeremy FYI that N2D machines no longer support the west zone nor the p100 GPU. You might want to update your documentation. @rachel FYI too.

After reading some GCP docs I realized that N2D machines are in beta and they are no longer supported in the west zone + they no longer support the p100 GPU.

I got the following setup to work which has a little more memory than the recommended setup but has the same GPU as recommended.

export IMAGE_FAMILY="pytorch-latest-gpu" 
export ZONE="us-west1-b"
export INSTANCE_NAME="my-fastai-instance"
export INSTANCE_TYPE="n1-highmem-16" # It seems like the N2D machines are in beta and are no longer available in all zones + not working with p100 anymore.

gcloud compute instances create $INSTANCE_NAME \
        --zone=$ZONE \
        --image-family=$IMAGE_FAMILY \
        --image-project=deeplearning-platform-release \
        --maintenance-policy=TERMINATE \
        --accelerator="type=nvidia-tesla-p100,count=1" \
        --machine-type=$INSTANCE_TYPE \
        --boot-disk-size=200GB \
        --metadata="install-nvidia-driver=True" \
        #--preemptible # Don’t use preemptible as it gave me issues before; described in this thread too.

0tist · April 14, 2020, 1:10pm

Hey, I recently moved to GCP from paperspace

I’m getting this error, though i opted for the passphrase, I have no clue what the problem is, I’m getting the key from the compute engine --> metadata --> SSH,
I have also tried getting it from the /Users/<sys_username>/.ssh/google_compute_engine.pub,
Need Help!

init_27 · April 14, 2020, 1:51pm

@steef I believe the OP wasn’t a wiki, I’ve fixed that. Could you please add these to the top post maybe under a new section?

You could also open a PR for the GCP setup instructions if you’d like.

Thank you!

AndreaPi · April 16, 2020, 1:27pm

Ooooh I see that @steef post already answered my question about which machine to use (-: so I deleted the original post. Just for curiosity, since I’m located in Europe, would it make more sense to use us-west1-b (the recommended zone for GCP) or an European one, e.g., europe-west4-b?

PS I can add these info to the OP wiki post, if @steef would rather not do that.

AndreaPi · April 16, 2020, 1:39pm

@init_27 I don’t want to steal @steef thunder but just for my own understanding: this is the doc one should update with a PR, right?

github.com

fastai/course-v3/blob/master/docs/start_gcp.md

---
title: GCP
sidebar: home_sidebar
---

# Welcome to GCP!

<img alt="" src="/images/gcp/logo.jpg" class="screenshot">

This guide explains how to set up Google Cloud Platform (GCP) to use PyTorch 1.0.0 and fastai 1.0.2. At the end of this tutorial you will be able to use both in a GPU-enabled Jupyter Notebook environment.

If you are returning to work and have previously completed the steps below, please go to the [returning to work](https://course.fast.ai/update_gcp.html) section.

## Pricing

GCP charges separately for the base CPU platform, and the GPU. We have two recommended configurations, *standard*, and *budget*. Note that all prices assume you are using *preemptible instances* (see below for details).

### Standard Compute

The base platform we suggest is called `n2d-highmem-8`, and costs $0.11 per hour. Attaching a T4 GPU costs $0.11 per hour so both together amount to **$0.22 per hour**.

This file has been truncated. show original

and these are the PR instructions for this repo:

github.com

fastai/course-v3/blob/master/docs/pr_tutorial.md

# How to make a Pull Request(PR)
For this tutorial we are going to use GitHub desktop, github.com, and a text editor of your choice.

## 1. **GitHub**
Make sure you are logged in to your GitHub account. ([sign up](https://github.com/join?source=header-home) or [sign in](https://github.com/login?return_to=%2Fjoin%3Fsource%3Dheader-home) )

## 2. **GitHub Desktop**
Download([windows](https://desktop.github.com) or [macOS](https://central.github.com/deployments/desktop/desktop/latest/darwin)) or open GitHub Desktop

## 3. **Tutorials** 
If you are new to GitHub, a good place to start is GitHub's ["Hello World" tutorial](https://guides.github.com/activities/hello-world/) it cover some of the features/terms you are going to need to know([What is GitHub?](https://guides.github.com/activities/hello-world/#what), [Create a Repository(Repo](https://guides.github.com/activities/hello-world/#repository), [Create a Branch](https://guides.github.com/activities/hello-world/#branch), [Make a Commit](https://guides.github.com/activities/hello-world/#commit), [Open a Pull Request(PR)](https://guides.github.com/activities/hello-world/#pr), [Merge Pull Request](https://guides.github.com/activities/hello-world/#merge)). Some other options are [hubspot's git-and-github-tutorial-for-beginners](https://product.hubspot.com/blog/git-and-github-tutorial-for-beginners), or [github-flow](https://guides.github.com/introduction/flow/)
  
## 4. **Clone & Fork** 
I chose to [Clone](https://help.github.com/articles/cloning-a-repository/) as well as fork the fastai GitHub repo. The reason I chose to do this is I could use the official GitHub repo for reading/class-work, and use the forked one to make my proposed change(s) to the library. 

## 5. **Clone It**
![clone|690x338](images\pr_tutorial\clone.png)
To me cloning a repo is like borrowing a library book, it is possible to make changes/commits(write in it) when it is in your possession, but changes you make could cause problems/errors(conflicts) since you don't have permission to make changes/updates. The changes you make will only be on your local copy, and if you try to return-it/update-it, it might cause issues. Since most library and companies have a code review process, you can't create a pull request(PR) from a cloned copy. You need to do that from a Forked copy.

## 6. **Fork It**

This file has been truncated. show original

Note that I’m pointing to v3, because the v4 repository doesn’t include a docs folder atm.

init_27 · April 16, 2020, 10:01pm

I believe this is the one

AndreaPi · April 19, 2020, 9:58pm

I edited the OP with the instructions by @gautam_e and @steef . Have a look and let me know what you think about it!

steef · April 24, 2020, 2:02am

Thanks @AndreaPi & @init_27! Not stealing my thunder at all – just glad my findings were useful for others too

alexxcollins21 · April 30, 2020, 9:00pm

Hello all - wondering if anyone more proficient with GCP can help:

I’m getting an error saying that us-west1-b is out of resource and I can’t start my fastai instance. Error message suggests I try a different zone, but any other zones I try that offer an Nvidia Tesla K80 work - I just get another error because I created my instance on the us-west1-b zone.

Is anyone aware of a quick fix? Or do I just have to wait for resources to become available?

Starting instance(s) alex-fastai-instance...failed.  
ERROR: (gcloud.compute.instances.start) The zone 'projects/fastai-alex-2020/zones/us-west1-b' 
does not have enough resources available to fulfill the request.  Try a different zone, or try again later.

When I try a different zone I get this:

$ gcloud compute instances start alex-fastai-instance --zone=us-central1-c
ERROR: (gcloud.compute.instances.start) HTTPError 404: 
The resource 'projects/fastai-alex-2020/zones/us-central1-c/instances/alex-fastai-instance' was not found

Many thanks,

Alex

florianl · May 2, 2020, 8:26pm

Did you create a preemptible imstance? With non-preemptible instances i had this Problem once but after a second try to start the instance it worked.

alexxcollins21 · May 3, 2020, 5:27pm

I created a non-preemptible instance (and preemptibility is still switched off) so hadn’t expected this. In my case I tried several times over about ten minutes and then spent some time going through Python for data analysis (Wes McKinney) which was definitely worth doing! It worked when I returned a couple of hours later.

Second “problem” I have with my GC account is that I’m actually being charged to use it. I had expected everything to be free on my trial so I’ll have to investigate that. Could be that I’ve used US located VM instance when I’m in Europe. I can’t remember the billing details well enough so it’s on the to-do list to solve at a later date…

0tist · May 10, 2020, 4:54pm

I have been trying to access my VM instance for a past couple of days, I’m stuck on the same error after (1)waiting patiently, (2)creating a new instance with a snapshot of my previous instance but with a different region
The zone 'projects/fast-ai-jay/zones/us-west1-b' does not have enough resources available to fulfill the request. Try a different zone, or try again later.
I tried making a new instance again, but i got this error:
Create VM instance "my-fastai-instance-3" and its boot disk "my-fastai-instance-3" 1 hour ago: fast-ai-jay The zone 'projects/fast-ai-jay/zones/australia-southeast1-c' does not have enough resources available to fulfill the request. Try a different zone, or try again later.

JorgeBriones · May 10, 2020, 6:32pm

Try creating a preemptible instance, this is what worked for me on us-west1-b. I had been on a non-preemptible instance for the duration of this course and during mid week of lesson 8 I started getting the same type of errors you have seen. After many tries, I went back to us-west1-b and decided to add on preemptible which worked! Since last Wednesday, I haven’t had any issues so far in regards to the resources error or being pre-empted.

0tist · May 11, 2020, 8:55am

great, that worked for me !
Thanks @JorgeBriones

gautam_e · May 11, 2020, 9:09am

Perhaps we should change the top post to reflect the commands for a preemptible instance, since that appears to be the solutions to the problems a lot of people are having? What do you the others think? Like this message if you are in favour of this suggestion.

AndreaPi · May 12, 2020, 3:25pm

Ok! At some point in time multiple users including me had the opposite issue, i.e., preemptible would never work, while non-preemptible ones would work right from the start. But GCP is a stochastic environment so it’s entirely possible that now preemptible instances work better. If you want to edit the Wiki, I’m fine with that!