Update Amazon docs:

`t2.micro` works but recommend `t3.small`

.github/integration-test.py

@@ -29,8 +29,8 @@ def run_systemd_image(image_name, container_name, bootstrap_pip_spec):
         '--mount', 'type=bind,source=/sys/fs/cgroup,target=/sys/fs/cgroup',
         '--detach',
         '--name', container_name,
-        # This is the minimum VM size we support.
-        # If we change this, need to change all other references to this number.
+        # A bit less than 1GB to ensure TLJH runs on 1GB VMs.
+        # If this is changed all docs references to the required memory must be changed too.
         '--memory', '900m',
     ]
 

docs/install/amazon.rst

@@ -76,8 +76,9 @@ Let's create the server on which we can run JupyterHub.
    `Next: Configure Instance Details` in the lower right corner.
 
    Check out our guide on How To :ref:`howto/admin/resource-estimation` to help pick
-   how much Memory / CPU your server needs. You need to have at least **1GB** of
-   RAM. The smallest instance that can fit this is a **t3.small**.
+   how much Memory / CPU your server needs.
+   We recommend you use a server with at leat 2GB of RAM, such as a **t3.small**.
+   However, if you need to minimise costs you can use a server with **1GB** RAM such as a **t2.micro**, but performance will be limited.
 
    You may wish to consult the listing `here <https://www.ec2instances.info/>`_
    because it shows cost per hour. The **On Demand** price is the pertinent cost.