Anyone here using GCP behind institute proxy?
Actually I can’t ssh GCP from my local machine because my institute network doesn’t listen on port 22 (default for process in GCP) so is there any way I can setup another port in GCP to listen from my local machine?
Looks like Google Cloud can not handle all the instances created on zone us-west1-b !
I have been getting an error in the past 2 days and I can’t start my instance !
Error : “The zone ‘projects/operating-land-216419/zones/us-west1-b’ does not have enough resources available to fulfill the request. Try a different zone, or try again later.”
Today I tried to start my machine like 20 times , all day long…
Also, I did a work that I’d like to publish , but I have no access to my notebook.
Very frustrating!
This link describes how to move the instance to a different zone. In my case, it raised the error that my instance is terminated and should be deleted (No idea why!). So I deleted it and set up a new instance in us-central1-c.
To list all instances and their status you can use this in command-line: gcloud compute instances list
Hey what do you replace $ZONE with? and I presume you type your instance name instead of $INSTACCE_NAME? Because if your ZONE is not similar to your default ones, it’s not connecting or atleast that was the case for me. Now it connects when ever I use the same default zone.
Yes, try changing your Firewall rules. You should find a firewall rule default-allow-ssh with tcp:22, try changing that to the port that your institute network listens to. Let me know if it works.
@sayko I’m sorry. Actually there wasn’t any fix i did, it’s only earlier i couldn’t able to find course-v3 folder may be they were updating something. But now it’s back to normal ( i mean now it’s showing course-v3 and anyone can do git pull) back to normal.
Finally after lots of struggle, lots of tring here and there, today I managed to setup my GCP:sweat_smile:. After installing google cloud sdk on my terminal, I was unable to run gcloud init. I found that gcloud is actually is an executable file so I changed the directory to bin where gcloud executable file is located and ran all the gcloud commands prefixing ./ i.e. ./gcloud init. With this minor modification I followed all the instructions as mentioned in fastai’s gcp guide and it worked. When I returned to work, I faced a problem similar to this https://forums.fast.ai/t/platform-gcp/27375/99?u=riteshc6 post and thanks to @sandeepsign for this post it helped.
P.S : I’m newbie to cloud computing so I just managed to get it done somehow. I don’t really understand what goes under the hood
Just as anecdotal proof: preemptible instances actually do get ‘preempted’ (mine 33% of the time as this was my 3rd time using it). It takes a while to understand that this time it is not just jupyter acting up but really the instance. You will find this in the logs of the VM:
This is very annoying if it happens after 4 hours of training (without any custom model saving callback inbetween, that’s my next todo… any suggestions?), money and time wasted…
When I stop my instance, I get this error in my terminal (gcloud.compute.ssh) [/usr/bin/ssh] exited with return code [255]
Any idea what it is, if it’s a problem? And what needs to done to avoid that?