The ec2imds python package is a module and a cli tool for retrieving AWS EC2 instance meta data and user data from the Instance Meta Data Service endpoint. It is an improved version of ec2metadata.

The cli tool prints data in JSON formate to stdout. The module consists of "low-level" utils for traversing the IMDS directory.

pip install ec2imds

On an EC2 instance, run the module.

# Get everything
python -m ec2imds

# Download user data(treated as binary)
python -m ec2imds -U > userdata

# Get specific categories
python -m ec2imds meta-data/instance-id meta-data/public-ipv4

# Get categories matching the regex
python -m ec2imds -E 'meta-data/placement/.*'

# List all implemented categories
python -m ec2imds -L

For more info on usage, run python -m ec2imds -h.

The cli tool can be used from a shell script in conjunction with jq.

ec2imds uses IMDS v2 only. The IMDS v2 mandates the use of token. IMDSWrapper provides callbacks for saving and loading tokens when sending requests to the IMDS endpoint. Incorporate the callbacks to suit your use case. You can use IMDSWrapper without the callbacks, but you may face issues from token exhaustion in the long run. Following is a simple example from the cli tool.

import datetime
import json
import os

from ec2imds import *

def on_new_token (
		token: str,
		expiry: datetime.datetime,
		endpoint: tuple[str, int]):
	# save the new token
	doc = {
		"token": token,
		"expiry": expiry.isoformat()
	}

	fp = rp.t + os.path.sep + IMDSWrapper.construct_endpoint_str(endpoint)
	os.makedirs(rp.t, 0o700, True)

	saved_umask = os.umask(0o077)
	with open(fp, "w") as f:
		json.dump(doc, f)
	os.umask(saved_umask)

def on_load_token (endpoint: tuple[str, int]) -> tuple[str, datetime.datetime]:
	fp = rp.t + os.path.sep + IMDSWrapper.construct_endpoint_str(endpoint)
	try:
		with open(fp, "rb") as f:
			t = json.load(f)
			return ( t["token"], datetime.datetime.fromisoformat(t["expiry"]) )
	except (FileNotFoundError, json.JSONDecodeError):
		return

Where

• rp.t is the path to the token directory.

In the token directory, a file will be created to save the taken value and its expiry for each IMDS endpoints, 169.254.169.254:80 and [fd00:ec2::254]:80.

Now, functions for fetching categories are defined in IMDSWrapper.dir_dict.

# Instantiate and set up
w = IMDSWrapper()
w.on_new_token = on_new_token
w.on_load_token = on_load_token

# Get all meta data
all_meta = w.all()

# Get user data
try:
	with (w.open_userdata() as u, open("file", "wb") as f):
		while f.write(w.read(4096)): pass
except:
	...

# Get specific meta data
instance_id = w.dir_dict["meta-data/instance-id"].func()

# Cycle through
for k, v in w.dir_dict.items():
	if k.startswith("meta-data/placement/"):
		print(v.func())

Note that for requests ended in 404, the functions return None.

The module does not define any classes for returned meta data. The functions always return dict for directory data and str, int, bool ... for scalar values. The module is, per se, a low-level util for fetching data from the IMDS.

See doc/ipv6only.md.