[lxc-users] AWS EC2: timeout connecting to instance metadata webserver (169.254.169.254) for *some* URLs (when connecting from a LXD container)

[Tomasz Chmielewski]

On 2020-11-19 00:07, Tomasz Chmielewski wrote:
> On 2020-11-18 23:50, Tomasz Chmielewski wrote:
>> That's a weird one!
>> 
>> In AWS, there is a concept of "instance metadata" - a webserver which
>> lets you fetch some instance metadata using http:
>> 
>> https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/instancedata-data-retrieval.html
>> 
>> 
>> For example, you can run this (from both AWS/EC2 instance and LXD
>> container running inside a AWS/EC2 instance), it will return some
>> metadata:
>> 
>> curl -v http://169.254.169.254/latest/meta-data/
>> 
>> 
>> Now, some of these requests time out when executed from a LXD
>> container running inside a AWS/EC2 - but work perfectly from the very
>> same AWS/EC2 instance.
>> 
>> 
>> For example, this request works fine from AWS/EC2 instance (ignore the
>> output - HTTP connection works just fine):
>> 
>> root at aws-instance:~# curl -v http://169.254.169.254/latest/api/token
>> *   Trying 169.254.169.254...
>> * TCP_NODELAY set
>> * Connected to 169.254.169.254 (169.254.169.254) port 80 (#0)
>>> GET /latest/api/token HTTP/1.1
>>> Host: 169.254.169.254
>>> User-Agent: curl/7.58.0
>>> Accept: */*
>>> 
>> < HTTP/1.1 405 Not Allowed
>> < Allow: OPTIONS, PUT
>> < Content-Length: 0
>> < Date: Wed, 18 Nov 2020 22:41:46 GMT
>> < Server: EC2ws
>> < Connection: close
>> < Content-Type: text/plain
>> <
>> * Closing connection 0
>> 
>> 
>> However, when executed from within a LXD container running inside the
>> very same AWS/EC2 instance - it times out!
>> 
>> root at lxd-container:~# curl -v http://169.254.169.254/latest/api/token
>> *   Trying 169.254.169.254...
>> * TCP_NODELAY set
>> * Connected to 169.254.169.254 (169.254.169.254) port 80 (#0)
>>> GET /latest/api/token HTTP/1.1
>>> Host: 169.254.169.254
>>> User-Agent: curl/7.58.0
>>> Accept: */*
>>> 
>> 
>> 
>> Even more weirdly, these work inside the container:
>> 
>> curl -v http://169.254.169.254/latest/api/
>> curl -v http://169.254.169.254/latest/api/t
>> curl -v http://169.254.169.254/latest/api/to
>> curl -v http://169.254.169.254/latest/api/tok
>> curl -v http://169.254.169.254/latest/api/toke
>> 
>> 
>> And this times out:
>> 
>> curl -v http://169.254.169.254/latest/api/token
>> 
>> 
>> 
>> Does anyone know why? tcpdump doesn't give me many clues (TTL?).
> 
> A somewhat related post (with docker having a similar issue):
> 
> https://rtfm.co.ua/en/aws-eksctl-put-http-169-254-169-254-latest-api-token-net-http-request-canceled-2/
> 
> 
> But, I'm no closer in getting a similar workaround for LXD.

If someone's struggling with a similar issue - here is a fix:

aws ec2 modify-instance-metadata-options --instance-id i-abcdefghijklmn 
--http-put-response-hop-limit 2 --http-endpoint enabled


Tomasz Chmielewski
https://lxadm.com