Expose EC2 platform type in machine readable format from ds-identify

[Robert C Jennings]

I would like to see the EC2 platform information in /run/cloud-init/cloud.cfg (or someplace else sensible) so that we can capture AWS-specific data for bugs filed by apport (bug #1724623) without reproducing dscheck_Ec2() from ds-identify.

Also, if you have suggestions as to what cloud-init info should be appended to bug descriptions or attachments as part of bug #1724623, please comment there as the insight would be very helpful.

[Chad Smith (chad.smith)]

Minimally this is resolved as of cloud-init version 21.4 which surfaced a /run/cloud-init/cloud-id file which represents the unique cloud datasource type cloud-init detects:

cat /run/cloud-init/cloud-id will have the following values:
aws
azure
gce
openstack
lxd
nocloud

Note as well that cloud-init creates a symlink pointing to this file of the format:
   /run/cloud-init/cloud-id-<cloud_name>

This allows scripts to easily check presence of a cloud-specific file /run/cloud-init/cloud-id-aws to indicate what platform cloud-init detected.

Note that this value will be incorrect if someone installs NoCloud datasource seed files on an Ec2 vm in /var/lib/cloud/seed/nocloud-net/(user-data|meta-data) because cloud-init will detect the NoCloud datasource before Ec2 datasource in that case.

Additional sources of flat files that can provide more context for identity of the system/platform

would be /run/cloud-init/instance-data.json which is structured configuration attributes that cloud-init reads from the datasource it detects.

instance-data.json provide versioning of generalized config variables under the v1 key of instance-data.json

Here are some examples on various clouds (again note that any cloud could have a NoCloud seed file placed in the environment that would prevent cloud-init from detecting the underlying cloud platform config on those systems):

https://hackmd.io/aAJFerDRRNaoL0Q_vDRYBg

Keep in mind as well there are additional pieces of data that are cloud-specific which give you more details about the offering, instance size etc under a top-level key in instance-data.json "ds".

On EC2 you get something like the following under 'ds'

{
 "_doc": "EXPERIMENTAL: The structure and format of content scoped under the 'ds' key may change in subsequent releases of cloud-init.",
 "_metadata_api_version": "2021-03-23",
 "dynamic": {
  "instance-identity": {
   "document": {
    "accountId": "329910648999",
    "architecture": "x86_64",
    "availabilityZone": "us-east-2b",
    "billingProducts": null,
    "devpayProductCodes": null,
    "imageId": "ami-0283a57753b18025b",
    "instanceId": "i-0ecd5e14298b43fbf",
    "instanceType": "t2.micro",
    "kernelId": null,
    "marketplaceProductCodes": null,
    "pendingTime": "2022-12-16T21:13:36Z",
    "privateIp": "172.31.24.213",
    "ramdiskId": null,
    "region": "us-east-2",
    "version": "2017-09-30"
   },
   "pkcs7": [
    "MIAGCSqGSIb3DQEHAqCAMI..."
   ],
   "rsa2048": [
    "MIAGCS..."
   ],
   "signature": [
    "Qk99HHJxS..."
   ]
  }
 },
 "meta-data": {
  "ami-id": "ami-0283a57753b18025b",
  "ami-launch-index": "0",
  "ami-manifest-path": "(unknown)",
  "block-device-mapping": {
   "ami": "/dev/sda1",
   "ephemeral0": "sdb",
   "ephemeral1": "sdc",
   "root": "/dev/sda1"
  },
  "events": {
   "maintenance": {
    "history": "[]",
    "scheduled": "[]"
   }
  },
  "hostname": "ip-172-31-24-213.us-east-2.compute.internal",
  "identity-credentials": {
   "ec2": {
    "info": {
     "AccountId": "32991064...",
     "Code": "Success",
     "LastUpdated": "2022-12-16T21:13:57Z"
    }
   }
  },
  "instance-action": "none",
  "instance-id": "i-0ecd5e14298b43fbf",
  "instance-li...