HPCE / hpce-2017-cw5

1 stars 6 forks source link

g2.2xlarge instance not launching in aws #24

Closed crypto-perry closed 6 years ago

crypto-perry commented 6 years ago

screen shot 2017-11-15 at 11 30 09 pm I get that when I try to launch the instance. Anyone getting the same? I already followed their advice and asked to raise the limit but what if that doesn't happen in due time.

guigzzz commented 6 years ago

It doesn't seem we have access to GPU instances using our AWS educate account. At least when I try to launch an instance, I cannot select a GPU instance.

m8pple commented 6 years ago

I think they usually raise the limit reasonably quickly - if you've spent some time launching other AWS instances, then they usually either don't impose the limit. Assuming everyone tried to start a g2.2xlarge for CW3, they will usually have allowed it by now.

So this only happens for g2.2xlarge - the c4.8xlarge was ok for CW4?

They are generally suspicious of accounts that are created and then people immediately try to go for the bigger/more expensive instances, as it could be people who have got a stolen credit card or something. They are also sensitive to GPU instances, as people try to use throwaway or stolen accounts to do bit-coin mining.

davidpasztor commented 6 years ago

Even normal AWS accounts have a limit on the number of instances you can run simultaneously and the default limit for g2.2xlarge instances is set to 0 by default. You can request a limit increase in the Limits tab. Limits For me it took a few days for them to approve the limit increase.

m8pple commented 6 years ago

@guigzzz Is this for a standard AWS Account with AWS Educate credits (the $100 grant), or for an AWS Educate Starter Account (the limited one with $75).

You should be able to start GPU instances using the normal AWS account, but may need to request an increase to your account limits (which can take a couple of days).

m8pple commented 6 years ago

Ah, thanks @davidpasztor - I think when I first used AWS they were more generous, and you could start up to 1 g2.2xlarge by default as long as you'd been using it for a bit.

guigzzz commented 6 years ago

@m8pple I realise now that I was simply looking at the wrong region (London doesn't have gpu instances apparently). Once I set the region to something like North Virginia everything's fine (although I still need to request a limit increase on the g2.2xlarge instance).

this should help anyone else trying to request an increase.

crypto-perry commented 6 years ago

I got my limit increased to 3 in 2 days so people should start requesting early

JB515 commented 6 years ago

Interesting case here, my g2.2xlarge limit started at 5, which I still have, but I get a permission error when I try to launch one.

capture

m8pple commented 6 years ago

Odd - is that repeatable if you explicitly log out of AWS and back in again?

I've seen it in the context of IAM roles, but not when logged in as myself.

JB515 commented 6 years ago

Yes, logging in and out doesn't make any difference.

m8pple commented 6 years ago

Afraid I couldn't find any more information or suggestions on it - it might need to be escalated to EC2 support.

Presumably your partner(s) are able to launch them, so you're ok for now.