"Apoptosis Network (alternate)" by "Simon Cockell" on Flickr

A few weird issues in the networking on our custom AWS EKS Workers, and how we worked around them

For “reasons”, at work we run AWS Elastic Kubernetes Service (EKS) with our own custom-built workers. These workers are based on Alma Linux 9, instead of AWS’ preferred Amazon Linux 2023. We manage the deployment of these workers using AWS Auto-Scaling Groups.

Our unusal configuration of these nodes mean that we sometimes trip over configurations which are tricky to get support on from AWS (no criticism of their support team, if I was in their position, I wouldn’t want to try to provide support for a customer’s configuration that was so far outside the recommended configuration either!)

Over the past year, we’ve upgraded EKS1.23 to EKS1.27 and then on to EKS1.31, and we’ve stumbled over a few issues on the way. Here are a couple of notes on the subject, in case they help anyone else in their journey.

All three of the issues below were addressed by running an additional service on the worker nodes in a Systemd timed service which triggers every minute.

Incorrect routing for the 12th IP address onwards

Something the team found really early on (around EKS 1.18 or somewhere around there) was that the AWS VPC-CNI wasn’t managing the routing tables on the node properly. We raised an issue on the AWS VPC CNI (we were on CentOS 7 at the time) and although AWS said they’d fixed the issue, we currently need to patch the routing tables every minute on our nodes.

What happens?

When you get past the number of IP addresses that a single ENI can have (typically ~12), the AWS VPC-CNI will attach a second interface to the worker, and start adding new IP addresses to that. The VPC-CNI should setup routing for that second interface, but for some reason, in our case, it doesn’t. You can see this happens because the traffic will come in on the second ENI, eth1, but then try to exit the node on the first ENI, eth0, with a tcpdump, like this:

[root@test-i-01234567890abcdef ~]# tcpdump -i any host 192.0.2.123
tcpdump: data link type LINUX_SLL2
dropped privs to tcpdump
tcpdump: verbose output suppressed, use -v[v]... for full protocol decode
listening on any, link-type LINUX_SLL2 (Linux cooked v2), snapshot length 262144 bytes
09:38:07.331619 eth1  In  IP ip-192-168-1-100.eu-west-1.compute.internal.41856 > ip-192-0-2-123.eu-west-1.compute.internal.irdmi: Flags [S], seq 1128657991, win 64240, options [mss 1359,sackOK,TS val 2780916192 ecr 0,nop,wscale 7], length 0
09:38:07.331676 eni989c4ec4a56 Out IP ip-192-168-1-100.eu-west-1.compute.internal.41856 > ip-192-0-2-123.eu-west-1.compute.internal.irdmi: Flags [S], seq 1128657991, win 64240, options [mss 1359,sackOK,TS val 2780916192 ecr 0,nop,wscale 7], length 0
09:38:07.331696 eni989c4ec4a56 In  IP ip-192-0-2-123.eu-west-1.compute.internal.irdmi > ip-192-168-1-100.eu-west-1.compute.internal.41856: Flags [S.], seq 3367907264, ack 1128657992, win 26847, options [mss 8961,sackOK,TS val 1259768406 ecr 2780916192,nop,wscale 7], length 0
09:38:07.331702 eth0  Out IP ip-192-0-2-123.eu-west-1.compute.internal.irdmi > ip-192-168-1-100.eu-west-1.compute.internal.41856: Flags [S.], seq 3367907264, ack 1128657992, win 26847, options [mss 8961,sackOK,TS val 1259768406 ecr 2780916192,nop,wscale 7], length 0

The critical line here is the last one – it’s come in on eth1 and it’s going out of eth0. Another test here is to look at ip rule

[root@test-i-01234567890abcdef ~]# ip rule
0:	from all lookup local
512:	from all to 192.0.2.111 lookup main
512:	from all to 192.0.2.143 lookup main
512:	from all to 192.0.2.66 lookup main
512:	from all to 192.0.2.113 lookup main
512:	from all to 192.0.2.145 lookup main
512:	from all to 192.0.2.123 lookup main
512:	from all to 192.0.2.5 lookup main
512:	from all to 192.0.2.158 lookup main
512:	from all to 192.0.2.100 lookup main
512:	from all to 192.0.2.69 lookup main
512:	from all to 192.0.2.129 lookup main
1024:	from all fwmark 0x80/0x80 lookup main
1536:	from 192.0.2.123 lookup 2
32766:	from all lookup main
32767:	from all lookup default

Notice here that we have two entries from all to 192.0.2.123 lookup main and from 192.0.2.123 lookup 2. Let’s take a look at what lookup 2 gives us, in the routing table

[root@test-i-01234567890abcdef ~]# ip route show table 2
192.0.2.1 dev eth1 scope link

Fix the issue

This is pretty easy – we need to add a default route if one doesn’t already exist. Long before I got here, my boss created a script which first runs ip route show table main | grep default to get the gateway for that interface, then runs ip rule list, looks for each lookup <number> and finally runs ip route add to put the default route on that table, the same as on the main table.

ip route add default via "${GW}" dev "${INTERFACE}" table "${TABLE}"

Is this still needed?

I know when we upgraded our cluster from EKS1.23 to EKS1.27, this script was still needed. When I’ve just checked a worker running EKS1.31, after around 12 hours of running, and a second interface being up, it’s not been needed… so perhaps we can deprecate this script?

Dropping packets to the containers due to Martians

When we upgraded our cluster from EKS1.23 to EKS1.27 we also changed a lot of the infrastructure under the surface (AlmaLinux 9 from CentOS7, Containerd and Runc from Docker, CGroups v2 from CGroups v1, and so on). We also moved from using an AWS Elastic Load Balancer (ELB) or “Classic Load Balancer” to AWS Network Load Balancer (NLB).

Following the upgrade, we started seeing packets not arriving at our containers and the system logs on the node were showing a lot of martian source messages, particularly after we configured our NLB to forward original IP source addresses to the nodes.

What happens

One thing we noticed was that each time we added a new pod to the cluster, it added a new eni[0-9a-f]{11} interface, but the sysctl value for net.ipv4.conf.<interface>.rp_filter (return path filtering – basically, should we expect the traffic to be arriving at this interface for that source?) in sysctl was set to 1 or “Strict mode” where the source MUST be the coming from the best return path for the interface it arrived on. The AWS VPC-CNI is supposed to set this to 2 or “Loose mode” where the source must be reachable from any interface.

In this case you’d tell this because you’d see this in your system journal (assuming you’ve got net.ipv4.conf.all.log_martians=1 configured):

Dec 03 10:01:19 test-i-01234567890abcdef kernel: IPv4: martian source 192.168.1.100 from 192.0.2.123, on dev eth1

The net result is that packets would be dropped by the host at this point, and they’d never be received by the containers in the pods.

Fix the issue

This one is also pretty easy. We run sysctl -a and loop through any entries which match net.ipv4.conf.([^\.]+).rp_filter = (0|1) and then, if we find any, we run sysctl -w net.ipv4.conf.\1.rp_filter = 2 to set it to the correct value.

Is this still needed?

Yep, absolutely. As of our latest upgrade to EKS1.31, if this value isn’t set, then it will drop packets. VPC-CNI should be fixing this, but for some reason it doesn’t. And setting the conf.ipv4.all.rp_filter to 2 doesn’t seem to make a difference, which is contrary to the documentation in the relevant Kernel documentation.

After 12 IP addresses are assigned to a node, Kubernetes services stop working for some pods

This was pretty weird. When we upgraded to EKS1.31 on our smallest cluster we initially thought we had an issue with CoreDNS, in that it sometimes wouldn’t resolve IP addresses for services (DNS names for services inside the cluster are resolved by <servicename>.<namespace>.svc.cluster.local to an internal IP address for the cluster – in our case, in the range 172.20.0.0/16). We upgraded CoreDNS to the EKS1.31 recommended version, v1.11.3-eksbuild.2 and that seemed to fix things… until we upgraded our next largest cluster, and things REALLY went wrong, but only when we had increased to over 12 IP addresses assigned to the node.

You might see this as frequent restarts of a container, particularly if you’re reliant on another service to fulfil an init container or the liveness/readyness check.

What happens

EKS1.31 moves KubeProxy from iptables or ipvs mode to nftables – a shift we had to make internally as AlmaLinux 9 no longer supports iptables mode, and ipvs is often quite flaky, especially when you have a lot of pod movements.

With a single interface and up to 11 IP addresses assigned to that interface, everything runs fine, but the moment we move to that second interface, much like in the first case above, we start seeing those pods attached to the second+ interface being unable to resolve service addresses. On further investigation, doing a dig from a container inside that pod to the service address of the CoreDNS service 172.20.0.10 would timeout, but a dig against the actual pod address 192.0.2.53 would return a valid response.

Under the surface, on each worker, KubeProxy adds a rule to nftables to say “if you try and reach 172.20.0.10, please instead direct it to 192.0.2.53”. As the containers fluctuate inside the cluster, KubeProxy is constantly re-writing these rules. For whatever reason though, KubeProxy currently seems unable to determine that a second or subsequent interface has been added, and so these rules are not applied to the pods attached to that interface…. or at least, that’s what it looks like!

Fix the issue

In this case, we wrote a separate script which was also triggered every minute. This script looks to see if the interfaces have changed by running ip link and looking for any interfaces called eth[0-9]+ which have changed, and then if it has, it runs crictl pods (which lists all the running pods in Containerd), looks for the Pod ID of KubeProxy, and then runs crictl stopp <podID> [1] and crictl rmp <podID> [1] to stop and remove the pod, forcing kubelet to restart the KubeProxy on the node.

[1] Yes, they aren’t typos, stopp means “stop the pod” and rmp means “remove the pod”, and these are different to stop and rm which relate to the container.

Is this still needed?

As this was what I was working on all-day yesterday, yep, I’d say so 😊 – in all seriousness though, if this hadn’t been a high-priority issue on the cluster, I might have tried to upgrade the AWS VPC-CNI and KubeProxy add-ons to a later version, to see if the issue was resolved, but at this time, we haven’t done that, so maybe I’ll issue a retraction later 😂

Featured image is “Apoptosis Network (alternate)” by “Simon Cockell” on Flickr and is released under a CC-BY license.

A padlock and chain on a rusted gate

Using #NetworkFirewall and #Route53 #DNS #Firewall to protect a private subnet’s egress traffic in #AWS

I wrote this post in January 2023, and it’s been languishing in my Drafts folder since then. I’ve had a look through it, and I can’t see any glaring reasons why I didn’t publish it so… it’s published… Enjoy 😁

If you’ve ever built a private subnet in AWS, you know it can be a bit tricky to get updates from the Internet – you end up having a NAT gateway or a self-managed proxy, and you can never be 100% certain that the egress traffic isn’t going somewhere you don’t want it to.

In this case, I wanted to ensure that outbound HTTPS traffic was being blocked if the SNI didn’t explicitly show the DNS name I wanted to permit through, and also, I only wanted specific DNS names to resolve. To do this, I used AWS Network Firewall and Route 53 DNS Firewall.

I’ve written this blog post, and followed along with this, I’ve created a set of terraform files to represent the steps I’ve taken.

The Setup

Let’s start this story from a simple VPC with three private subnets for my compute resources, and three private subnets for the VPC Endpoints for Systems Manager (SSM).

Here’s our network diagram, with the three subnets containing the VPC Endpoints at the top, and the three instances at the bottom.

I’ve created a tag in my Github repo at this “pre-changes” state, called step 1.

At this point, none of those instances can reach anything outside the network, with the exception of the SSM environment. So, we can’t install any packages, we can’t get data from outside the network or anything similar.

Getting Protected Internet Access

In order to get internet access, we need to add 4 things;

  1. An internet gateway
  2. A NAT gateway in each AZ
  3. Which needs three new subnets
  4. And three Elastic IP addresses
  5. Route tables in all the subnets

To clarify, a NAT gateway acts like a DSL router. It hides the source IP address of outbound traffic behind a single, public IP address (using an Elastic IP from AWS), and routes any return traffic back to wherever that traffic came from. To reduce inter-AZ data transfer rates, I’m putting one in each AZ, but if there’s not a lot of outbound traffic or the outbound traffic isn’t critical enough to require resiliency, this could all be centralised to a single NAT gateway. To put a NAT gateway in each AZ, you need a subnet in each AZ, and to get out to the internet (by whatever means you have), you need an internet gateway and route tables for how to reach the NAT and internet gateways.

We also should probably add, at this point, four additional things.

  1. The Network Firewall
  2. Subnets for the Firewall interfaces
  3. Stateless Policy
  4. Stateful Policy

The Network Firewall acts like a single appliance, and uses a Gateway Load Balancer to present an interface into each of the availability zones. It has a stateless policy (which is very fast, but needs to address both inbound and outbound traffic flows) to do IP and Port based filtering (referred to as “Layer 3” filtering) and then specific traffic can be passed into a stateful policy (which is slower) to do packet and flow inspection.

In this case, I only want outbound HTTPS traffic to be passed, so my stateless rule group is quite simple;

  • VPC range on any port → Internet on TCP/443; pass to Stateful rule groups
  • Internet on TCP/443 → VPC range on any port; pass to Stateful rule groups

I have two stateful rule groups, one is defined to just allow access out to example.com and any relevant subdomains, using the “Domain List” stateful policy item. The other allows access to example.org and any relevant subdomains, using a Suricata stateful policy item, to show the more flexible alternative route. (Suricata has lots more filters than just the SNI value, you can check for specific SSH versions, Kerberos CNAMEs, SNMP versions, etc. You can also add per-rule logging this way, which you can’t with the Domain List route).

These are added to the firewall policy, which also defines that if a rule doesn’t match a stateless rule group, or an established flow doesn’t match a stateful rule group, then it should be dropped.

New network diagram with more subnets and objects, but essentially, as described in the paragraphs above. Traffic flows from the instances either down towards the internet, or up towards the VPCe.

I’ve created a tag in my Github repo at this state, with the firewall, NAT Gateway and Internet Gateway, called step 2.

So far, so good… but why let our users even try to resolve the DNS name of a host they’re not permitted to reach. Let’s turn on DNS Firewalling too.

Turning on Route 53 DNS Firewall

You’ll notice that in the AWS Network Firewall, I didn’t let DNS out of the network. This is because, by default, AWS enables Route 53 as it’s local resolver. This lives on the “.2” address of the VPC, so in my example environment, this would be 198.18.0.2. Because it’s a local resolver, it won’t cross the Firewall exiting to the internet. You can also make Route 53 use your own DNS servers for specific DNS resolution (for example, if you’re running an Active Directory service inside your network).

Any Network Security Response team members you have working with you would appreciate it if you’d turn on DNS Logging at this point, so I’ll do it too!

In March 2021, AWS announced “Route 53 DNS Firewall”, which allow this DNS resolver to rewrite responses, or even to completely deny the existence of a DNS record. With this in mind, I’m going to add some custom DNS rules.

The first thing I want to do is to only permit traffic to my specific list of DNS names – example.org, example.com and their subdomains. DNS quite likes to terminate DNS names with a dot, signifying it shouldn’t try to resolve any higher up the chain, so I’m going to make a “permitted domains” DNS list;

example.com.
example.org.
*.example.com.
*.example.org.

Nice and simple! Except, this also stops me from being able to access the instances over SSM, so I’ll create a separate “VPCe” DNS list:

ssm.ex-ample-1.amazonaws.com.
*.ssm.ex-ample-1.amazonaws.com.
ssmmessages.ex-ample-1.amazonaws.com.
*.ssmmessages.ex-ample-1.amazonaws.com.
ec2messages.ex-ample-1.amazonaws.com.
*.ec2messages.ex-ample-1.amazonaws.com.

Next I create a “default deny” DNS list:

*.

And then build a DNS Firewall Policy which allows access to the “permitted domains”, “VPCe” lists, but blocks resolution of any “default deny” entries.

I’ve created a tag in my Github repo at this state, with the Route 53 DNS Firewall configured, called step 3.

In conclusion…

So there we have it. While the network is not “secure” (there’s still a few gaps here) it’s certainly MUCH more secure than it was, and it certainly would take a lot more work for anyone with malicious intent to get your content out.

Feel free to have a poke around, and leave comments below if this has helped or is of interest!

A photo of a door with the focus on the handle which has a lock in the centre of the knob. The lock has a key in it, with a bunch of keys dangling from the central ring.

Using direnv with terraform, terragrunt, saml2aws, SOPS and AWS KMS

In my current project I am often working with Infrastructure as Code (IoC) in the form of Terraform and Terragrunt files. Before I joined the team a decision was made to use SOPS from Mozilla, and this is encrypted with an AWS KMS key. You can only access specific roles using the SAML2AWS credentials, and I won’t be explaining how to set that part up, as that is highly dependant on your SAML provider.

While much of our environment uses AWS, we do have a small presence hosted on-prem, using a hypervisor service. I’ll demonstrate this with Proxmox, as this is something that I also use personally :)

Firstly, make sure you have all of the above tools installed! For one stage, you’ll also require yq to be installed. Ensure you’ve got your shell hook setup for direnv as we’ll need this later too.

Late edit 2023-07-03: There was a bug in v0.22.0 of the terraform which didn’t recognise the environment variables prefixed PROXMOX_VE_ – a workaround by using TF_VAR_PROXMOX_VE and a variable "PROXMOX_VE_" {} block in the Terraform code was put in place for the inital publication of this post. The bug was fixed in 0.23.0 which this post now uses instead, and so as a result the use of TF_VAR_ prefixed variables was removed too.

Set up AWS Vault

AWS KMS

AWS Key Management Service (KMS) is a service which generates and makes available encryption keys, backed by the AWS service. There are *lots* of ways to cut that particular cake, but let’s do this a quick and easy way… terraform

variable "name" {
  default = "SOPS"
  type    = string
}
resource "aws_kms_key" "this" {
  tags                     = {
    Name : var.name,
    Owner : "Admins"
  }
  key_usage                = "ENCRYPT_DECRYPT"
  customer_master_key_spec = "SYMMETRIC_DEFAULT"
  deletion_window_in_days  = 30
  is_enabled               = true
  enable_key_rotation      = false
  policy                   = <<EOF
{
  "Version": "2012-10-17",
  "Id": "key-default-1",
  "Statement": [
    {
      "Sid": "Root Access",
      "Effect": "Allow",
      "Principal": {
        "AWS": "arn:aws:iam::${get_aws_account_id()}:root"
      },
      "Action": "kms:*",
      "Resource": "*"
    },
    {
      "Sid": "Estate Admin Access",
      "Effect": "Allow",
      "Principal": {
        "AWS": "arn:aws:iam::${get_aws_account_id()}:role/estateadmins"
      },
      "Action": [
        "kms:Describe*",
        "kms:List*",
        "kms:Get*",
        "kms:Encrypt*"
      ],
      "Resource": "*"
    }
  ]
}
EOF
}

resource "aws_kms_alias" "this" {
  target_key_id = aws_kms_key.this.key_id
  name          = "alias/${var.name}"
}

output "key" {
  value = aws_kms_alias.this.arn
}

After running this, let’s assume that we get an output for the “key” value of:

arn:aws:kms:us-east-1:123456789012:alias/main

Setup Sops

In your terragrunt tree, create a file called .sops.yaml, which contains:

---
creation_rules:
  - kms: arn:aws:kms:us-east-1:123456789012:alias/main

And a file called secrets.enc.yaml which contains:

---
PROXMOX_VE_USERNAME: root@pam
PROXMOX_VE_PASSWORD: deadb33f@2023

Test that your KMS works by assuming your IAM role via SAML2AWS like this:

$ saml2aws login --skip-prompt --quiet
$ saml2aws exec -- sops --verbose --encrypt --in-place secrets.enc.yaml
[AWSKMS]	 INFO[0000] Encryption succeeded                          arn="arn:aws:kms:us-east-1:123456789012:alias/main"
[CMD]		 INFO[0000] File written successfully

Setup direnv

Outside your tree, in ~/.config/direnv/lib create a file called use_sops.sh (does not need to be chmod +x or chmod 755!) containing this:

# Based on https://github.com/direnv/direnv/wiki/Sops
use_sops() {
    local path=${1:-$PWD/secrets.enc.yaml}
    if [ -e "$path" ]
    then
        if grep -q -E '^sops:' "$path"
        then
            eval "$(sops --decrypt --output-type dotenv "$path" 2>/dev/null | direnv dotenv bash /dev/stdin || false)"
        else
            if [ -n "$(command -v yq)" ]
            then
                eval "$(yq eval --output-format props "$path" | direnv dotenv bash /dev/stdin)"
                export SOPS_WARNING="unencrypted $path"
            fi
        fi
    fi
    watch_file "$path"
}

There are two key lines here, the first of which is:

eval "$(sops -d --output-type dotenv "$path" 2>/dev/null | direnv dotenv bash /dev/stdin || false)"

This line asks sops to decrypt the secrets file, using the “dotenv” output type, however, the dotenv format looks like this:

some_key = "some value"

So, as a result, we then pass that value to direnv and ask it to rewrite it in the format it expects, which looks like this:

export some_key="some value"

The second key line is this:

eval "$(yq eval --output-format props "$path" | direnv dotenv bash /dev/stdin)"

This asks yq to parse the secrets file, using the “props” formatter, which results in lines just like the dotenv output we saw above.

However, because we used yq to parse the file, it means that we know this file isn’t encrypted, so we also add an extra export value:

export SOPS_WARNING="unencrypted $path"

This can be picked up as part of your shell prompt to put a warning in! Anyway… let’s move on.

Now that you have your reusable library file, we now configure the direnv file, .envrc for the root of your proxmox cluster:

use sops

Oh, ok, that was simple. You can add several files here if you wish, like this:

use sops file1.enc.yaml
use sops file2.enc.yml
use sops ~/.core_sops

But, we don’t need that right now!

Open your shell in that window, and you’ll get this warning:

direnv: error /path/to/demo/.envrc is blocked. Run `direnv allow` to approve its content

So, let’s do that!

$ direnv allow
direnv: loading /path/to/demo/.envrc
direnv: using sops
direnv: export +PROXMOX_VE_USERNAME +PROXMOX_VE_PASSWORD
$

So far, so good… but wait, you’ve authenticated to your SAML access to AWS. Let’s close that shell, and go back in again

$ cd /path/to/demo
direnv: loading /path/to/demo/.envrc
direnv: using sops
$

Ah, now we don’t have our values exported. That’s what we wanted!

What now?!

Configuring the details of the proxmox cluster

We have our .envrc file which provides our credentials (let’s pretend we’re using a shared set of credentials across all the boxes), but now we need to setup access to each of the boxes.

Let’s make our two cluster directories;

mkdir cluster_01
mkdir cluster_02

And in each of these clusters, we need to put an .envrc file with the right IP address in. This needs to check up the tree for any credentials we may have already loaded:

source_env "$(find_up ../.envrc)"
export PROXMOX_VE_ENDPOINT="https://192.0.2.1:8006" # Documentation IP address for the first cluster - change for the second cluster.

The first line works up the tree, looking for a parent .envrc file to inject, and then, with the second line, adds the Proxmox API endpoint to the end of that chain. When we run direnv allow (having logged back into our saml2aws session), we get this:

$ direnv allow
direnv: loading /path/to/demo/cluster_01/.envrc
direnv: loading /path/to/demo/.envrc
direnv: using sops
direnv: export +PROXMOX_VE_ENDPOINT +PROXMOX_VE_USERNAME +PROXMOX_VE_PASSWORD
$

Great, now we can setup the connection to the cluster in the terragrunt file!

Set up Terragrunt

In /path/to/demo/terragrunt.hcl put this:

remote_state {
  backend = "s3"
  config  = {
    encrypt                = true
    bucket                 = "example-inc-terraform-state"
    key                    = "${path_relative_to_include()}/terraform.tfstate"
    region                 = "us-east-1"
    dynamodb_table         = "example-inc-terraform-state-lock"
    skip_bucket_versioning = false
  }
}
generate "providers" {
  path      = "providers.tf"
  if_exists = "overwrite"
  contents  = <<EOF
terraform {
  required_providers {
    proxmox = {
      source = "bpg/proxmox"
      version = "0.23.0"
    }
  }
}

provider "proxmox" {
  insecure = true
}
EOF
}

Then in the cluster_01 directory, create a directory for the code you want to run (e.g. create a VLAN might be called “VLANs/30/“) and put in it this terragrunt.hcl

terraform {
  source = "${get_terragrunt_dir()}/../../../terraform-module-network//vlan"
  # source = "git@github.com:YourProject/terraform-module-network//vlan?ref=production"
}

include {
  path = find_in_parent_folders()
}

inputs = {
  vlan_tag    = 30
  description = "VLAN30"
}

This assumes you have a terraform directory called terraform-module-network/vlan in a particular place in your tree or even better, a module in your git repo, which uses the input values you’ve provided.

That double slash in the source line isn’t a typo either – this is the point in that tree that Terragrunt will copy into the directory to run terraform from too.

A quick note about includes and provider blocks

The other key thing is that the “include” block loads the values from the first matching terragrunt.hcl file in the parent directories, which in this case is the one which defined the providers block. You can’t include multiple different parent files, and you can’t have multiple generate blocks either.

Running it all together!

Now we have all our depending files, let’s run it!

user@host:~$ cd test
direnv: loading ~/test/.envrc
direnv: using sops
user@host:~/test$ saml2aws login --skip-prompt --quiet ; saml2aws exec -- bash
direnv: loading ~/test/.envrc
direnv: using sops
direnv: export +PROXMOX_VE_USERNAME +PROXMOX_VE_PASSWORD
user@host:~/test$ cd cluster_01/VLANs/30
direnv: loading ~/test/cluster_01/.envrc
direnv: loading ~/test/.envrc
direnv: using sops
direnv: export +PROXMOX_VE_ENDPOINT +PROXMOX_VE_USERNAME +PROXMOX_VE_PASSWORD
user@host:~/test/cluster_01/VLANs/30$ terragrunt apply
data.proxmox_virtual_environment_nodes.available_nodes: Reading...
data.proxmox_virtual_environment_nodes.available_nodes: Read complete after 0s [id=nodes]

Terraform used the selected providers to generate the following execution
plan. Resource actions are indicated with the following symbols:
  + create

Terraform will perform the following actions:

  # proxmox_virtual_environment_network_linux_bridge.this[0] will be created
  + resource "proxmox_virtual_environment_network_linux_bridge" "this" {
      + autostart  = true
      + comment    = "VLAN30"
      + id         = (known after apply)
      + mtu        = (known after apply)
      + name       = "vmbr30"
      + node_name  = "proxmox01"
      + ports      = [
          + "enp3s0.30",
        ]
      + vlan_aware = (known after apply)
    }

Plan: 1 to add, 0 to change, 0 to destroy.

Do you want to perform these actions?
  Terraform will perform the actions described above.
  Only 'yes' will be accepted to approve.

  Enter a value: yes
proxmox_virtual_environment_network_linux_bridge.this[0]: Creating...
proxmox_virtual_environment_network_linux_bridge.this[0]: Creation complete after 2s [id=proxmox01:vmbr30]
user@host:~/test/cluster_01/VLANs/30$

Winning!!

Featured image is “2018/365/1 Home is Where The Key Fits” by “Alan Levine” on Flickr and is released under a CC-0 license.

"2009.01.17 - UNKNOWN, Unknown" by "Adrian Clark" on Flickr

Creating tagged AWS EC2 resources (like Elastic IPs) with Ansible

This is a quick note, having stumbled over this one today.

Mostly these days, I’m used to using Terraform to create Elastic IP (EIP) items in AWS, and I can assign tags to them during creation. For various reasons in $Project I’m having to create my EIPs in Ansible.

To make this work, you can’t just create an EIP with tags (like you would in Terraform), instead what you need to do is to create the EIP and then tag it, like this:

  - name: Allocate a new elastic IP
    community.aws.ec2_eip:
      state: present
      in_vpc: true
      region: eu-west-1
    register: eip

  - name: Tag that resource
    amazon.aws.ec2_tag:
      region: eu-west-1
      resource: "{{ eip.allocation_id }}"
      state: present
      tags:
        Name: MyTag
    register: tag

Notice that we create a VPC associated EIP, and assign the allocation_id from the result of that module to the resource we want to tag.

How about if you’re trying to be a bit more complex?

Here I have a list of EIPs I want to create, and then I pass this into the ec2_eip module, like this:

- name: Create list of EIPs
  set_fact:
    region: eu-west-1
    eip_list:
    - demo-eip-1
    - demo-eip-2
    - demo-eip-3

  - name: Allocate new elastic IPs
    community.aws.ec2_eip:
      state: present
      in_vpc: true
      region: "{{ region }}"
    register: eip
    loop: "{{ eip_list | dict2items }}"
    loop_control:
      label: "{{ item.key }}"

  - name: Tag the EIPs
    amazon.aws.ec2_tag:
      region: "{{ item.invocation.module_args.region }}"
      resource: "{{ item.allocation_id }}"
      state: present
      tags:
        Name: "{{ item.item.key }}"
    register: tag
    loop: "{{ eip.results }}"
    loop_control:
      label: "{{ item.item.key }}"

So, in this instance we pass the list of EIP names we want to create as a list with the loop instruction. Now, at the point we create them, we don’t actually know what they’ll be called, but we’re naming them there because when we tag them, we get the “item” (from the loop) that was used to create the EIP. When we then tag the EIP, we can use some of the data that was returned from the ec2_eip module (region, EIP allocation ID and the name we used as the loop key). I’ve trimmed out the debug statements I created while writing this, but here’s what you get back from ec2_eip:

"eip": {
        "changed": true,
        "msg": "All items completed",
        "results": [
            {
                "allocation_id": "eipalloc-decafbaddeadbeef1",
                "ansible_loop_var": "item",
                "changed": true,
                "failed": false,
                "invocation": {
                    "module_args": {
                        "allow_reassociation": false,
                        "aws_access_key": null,
                        "aws_ca_bundle": null,
                        "aws_config": null,
                        "aws_secret_key": null,
                        "debug_botocore_endpoint_logs": false,
                        "device_id": null,
                        "ec2_url": null,
                        "in_vpc": true,
                        "private_ip_address": null,
                        "profile": null,
                        "public_ip": null,
                        "public_ipv4_pool": null,
                        "region": "eu-west-1",
                        "release_on_disassociation": false,
                        "reuse_existing_ip_allowed": false,
                        "security_token": null,
                        "state": "present",
                        "tag_name": null,
                        "tag_value": null,
                        "validate_certs": true,
                        "wait_timeout": null
                    }
                },
                "item": {
                    "key": "demo-eip-1",
                    "value": {}
                },
                "public_ip": "192.0.2.1"
            }
     ]
}

So, that’s what I’m doing next!

Featured image is “2009.01.17 – UNKNOWN, Unknown” by “Adrian Clark” on Flickr and is released under a CC-BY-ND license.

"pharmacy" by "Tim Evanson" on Flickr

AWX – The Gateway Drug to Ansible Tower

A love letter to Ansible Tower

I love Ansible… I mean, I really love Ansible. You can ask anyone, and they’ll tell you my first love is my wife, then my children… and then it’s Ansible.

OK, maybe it’s Open Source and then Ansible, but either way, Ansible is REALLY high up there.

But, while I love Ansible, I love what Ansible Tower brings to an environment. See, while you get to easily and quickly manage a fleet of machines with Ansible, Ansible Tower gives you the fine grained control over what you need to expose to your developers, your ops team, or even, in a fit of “what-did-you-just-do”-ness, your manager. (I should probably mention that Ansible Tower is actually part of a much larger portfolio of products, called Ansible Automation Platform, and there’s some hosted SaaS stuff that goes with it… but the bit I really want to talk about is Tower, so I’ll be talking about Tower and not Ansible Automation Platform. Sorry!)

Ansible Tower has a scheduling engine, so you can have a “Go” button, for deploying the latest software to your fleet, or just for the 11PM patching cycle. It has a credential store, so your teams can’t just quickly go and perform an undocumented quick fix on that “flaky” box – they need to do their changes via Ansible. And lastly, it has an inventory, so you can see that the last 5 jobs failed to deploy on that host, so maybe you’ve got a problem with it.

One thing that people don’t so much love to do, is to get a license to deploy Tower, particularly if they just want to quickly spin up a demonstration for some colleagues to show how much THEY love Ansible. And for those people, I present AWX.

The first hit is free

One of the glorious and beautiful things that RedHat did, when they bought Ansible, was to make the same assertion about the Ansible products that they make to the rest of their product line, which is… while they may sell a commercial product, underneath it will be an Open Source version of that product, and you can be part of developing and improving that version, to help improve the commercial product. Thus was released AWX.

Now, I hear the nay-sayers commenting, “but what if you have an issue with AWX at 2AM, how do you get support on that”… and to those people, I reply: “If you need support at 2AM for your box, AWX is not the tool for you – what you need is Tower.”… Um, I mean Ansible Automation Platform. However, Tower takes a bit more setting up than what I’d want to do for a quick demo, and it has a few more pre-requisites. ANYWAY, enough about dealing with the nay-sayers.

AWX is an application inside Docker containers. It’s split into three parts, the AWX Web container, which has the REST API. There’s also a PostgreSQL database inside there too, and one “Engine”, which is the separate container which gets playbooks from your version control system, asks for any dynamic inventories, and then runs those playbooks on your inventories.

I like running demos of Tower, using AWX, because it’s reasonably easy to get stood up, and it’s reasonably close to what Tower looks and behaves like (except for the logos)… and, well, it’s a good gateway to getting people interested in what Tower can do for them, without them having to pay (or spend time signing up for evaluation licenses) for the environment in the first place.

And what’s more, it can all be automated

Yes, folks, because AWX is just a set of docker containers (and an install script), and Ansible knows how to start Docker containers (and run an install script), I can add an Ansible playbook to my cloud-init script, Vagrantfile or, let’s face it, when things go really wrong, put it in a bash script for some poor keyboard jockey to install for you.

If you’re running a demo, and you don’t want to get a POC (proof of concept) or evaluation license for Ansible Tower, then the chances are you’re probably not running this on RedHat Enterprise Linux (RHEL) either. That’s OK, once you’ve sold the room on using Tower (by using AWX), you can sell them on using RHEL too. So, I’ll be focusing on using CentOS 8 instead. Partially because there’s a Vagrant box for CentOS 8, but also because I can also use CentOS 8 on AWS, where I can prove that the Ansible Script I’m putting into my Vagrantfile will also deploy nicely via Cloud-Init too. With a very small number of changes, this is likely to work on anything that runs Docker, so everything from Arch to Ubuntu… probably 😁

“OK then. How can you work this magic, eh?” I hear from the back of the room. OK, pipe down, nay-sayers.

First, install Ansible on your host. You just need to run dnf install -y ansible.

Next, you need to install Docker. This is a marked difference between AWX and Ansible Tower, as AWX is based on Docker, but Ansible Tower uses other magic to make it work. When you’re selling the benefits of Tower, note that it’s not a 1-for-1 match at this point, but it’s not a big issue. Fortunately, CentOS can install Docker Community edition quite easily. At this point, I’m swapping to using Ansible playbooks. At the end, I’ll drop a link to where you can get all this in one big blob… In fact, we’re likely to use it with our Cloud-Init deployment.

Aw yehr, here’s the good stuff

tasks:
- name: Update all packages
  dnf:
    name: "*"
    state: latest

- name: Add dependency for "yum config-manager"
  dnf:
    name: yum-utils
    state: present

- name: Add the Docker Repo
  shell: yum config-manager --add-repo https://download.docker.com/linux/centos/docker-ce.repo
  args:
    creates: /etc/yum.repos.d/docker-ce.repo
    warn: false

- name: Install Docker
  dnf:
    name:
    - docker-ce
    - docker-ce-cli
    - containerd.io
    state: present
  notify: Start Docker

That first stanza – update all packages? Well, that’s because containerd.io relies on a newer version of libseccomp, which hasn’t been built in the CentOS 8 Vagrantbox I’m using.

The next one? That ensures I can run yum config-manager to add a repo. I could use the copy module in Ansible to create the repo files so yum and/or dnf could use that instead, but… meh, this is a single line shell command.

And then we install the repo, and the docker-ce packages we require. We use the “notify” statement to trigger a handler call to start Docker, like this:

handlers:
- name: Start Docker
  systemd:
    name: docker
    state: started

Fab. We’ve got Docker. Now, let’s clone the AWX repo to our machine. Again, we’re doing this with Ansible, naturally :)

tasks:
- name: Clone AWX repo to local path
  git:
    repo: https://github.com/ansible/awx.git
    dest: /opt/awx

- name: Get latest AWX tag
  shell: |
    if [ $(git status -s | wc -l) -gt 0 ]
    then
      git stash >/dev/null 2>&1
    fi
    git fetch --tags && git describe --tags $(git rev-list --tags --max-count=1)
    if [ $(git stash list | wc -l) -gt 0 ]
    then
      git stash pop >/dev/null 2>&1
    fi
  args:
    chdir: /opt/awx
  register: latest_tag
  changed_when: false

- name: Use latest released version of AWX
  git:
    repo: https://github.com/ansible/awx.git
    dest: /opt/awx
    version: "{{ latest_tag.stdout }}"

OK, there’s a fair bit to get from this, but essentially, we clone the repo from Github, then ask (using a collection of git commands) for the latest released version (yes, I’ve been bitten by just using the head of “devel” before), and then we check out that released version.

Fab, now we can configure it.

tasks:
- name: Set or Read admin password
  set_fact:
    admin_password_was_generated: "{{ (admin_password is defined or lookup('env', 'admin_password') != '') | ternary(false, true) }}"
    admin_password: "{{ admin_password | default (lookup('env', 'admin_password') | default(lookup('password', 'pw.admin_password chars=ascii_letters,digits length=20'), true) ) }}"

- name: Configure AWX installer
  lineinfile:
    path: /opt/awx/installer/inventory
    regexp: "^#?{{ item.key }}="
    line: "{{ item.key }}={{ item.value }}"
  loop:
  - key: "awx_web_hostname"
    value: "{{ ansible_fqdn }}"
  - key: "pg_password"
    value: "{{ lookup('password', 'pw.pg_password chars=ascii_letters,digits length=20') }}"
  - key: "rabbitmq_password"
    value: "{{ lookup('password', 'pw.rabbitmq_password chars=ascii_letters,digits length=20') }}"
  - key: "rabbitmq_erlang_cookie"
    value: "{{ lookup('password', 'pw.rabbitmq_erlang_cookie chars=ascii_letters,digits length=20') }}"
  - key: "admin_password"
    value: "{{ admin_password }}"
  - key: "secret_key"
    value: "{{ lookup('password', 'pw.secret_key chars=ascii_letters,digits length=64') }}"
  - key: "create_preload_data"
    value: "False"
  loop_control:
    label: "{{ item.key }}"

If we don’t already have a password defined, then create one. We register the fact we’ve had to create one, as we’ll need to tell ourselves it once the build is finished.

After that, we set a collection of values into the installer – the hostname, passwords, secret keys and so on. It loops over a key/value pair, and passes these to a regular expression rewrite command, so at the end, we have the settings we want, without having to change this script between releases.

When this is all done, we execute the installer. I’ve seen this done two ways. In an ideal world, you’d throw this into an Ansible shell module, and get it to execute the install, but the problem with that is that the AWX install takes quite a while, so I’d much rather actually be able to see what’s going on… and so, instead, we exit our prepare script at this point, and drop back to the shell to run the installer. Let’s look at both options, and you can decide which one you want to do. In my script, I’m doing the first, but just because it’s a bit neater to have everything in one place.

- name: Run the AWX install.
  shell: ansible-playbook -i inventory install.yml
  args:
    chdir: /opt/awx/installer
cd /opt/awx/installer
ansible-playbook -i inventory install.yml

When this is done, you get a prepared environment, ready to access using the username admin and the password of … well, whatever you set admin_password to.

AWX takes a little while to stand up, so you might want to run this next Ansible stanza to see when it’s ready to go.

- name: Test access to AWX
  tower_user:
    tower_host: "http://{{ ansible_fqdn }}"
    tower_username: admin
    tower_password: "{{ admin_password }}"
    email: "admin@{{ ansible_fqdn }}"
    first_name: "admin"
    last_name: ""
    password: "{{ admin_password }}"
    username: admin
    superuser: yes
    auditor: no
  register: _result
  until: _result.failed == false
  retries: 240 # retry 240 times
  delay: 5 # pause for 5 sec between each try

The upshot to using that command there is that it sets the email address of the admin account to “admin@your.awx.example.org“, if the fully qualified domain name (FQDN) of your machine is your.awx.example.org.

Moving from the Theoretical to the Practical

Now we’ve got our playbook, let’s wrap this up in both a Vagrant Vagrantfile and a Terraform script, this means you can deploy it locally, to test something internally, and in “the cloud”.

To simplify things, and because the version of Ansible deployed on the Vagrant box isn’t the one I want to use, I am using a single “user-data.sh” script for both Vagrant and Terraform. Here that is:

#!/bin/bash
if [ -e "$(which yum)" ]
then
  yum install git python3-pip -y
  pip3 install ansible docker docker-compose
else
  echo "This script only supports CentOS right now."
  exit 1
fi

git clone https://gist.github.com/JonTheNiceGuy/024d72f970d6a1c6160a6e9c3e642e07 /tmp/Install_AWX
cd /tmp/Install_AWX
/usr/local/bin/ansible-playbook Install_AWX.yml

While they both have their differences, they both can execute a script once the machine has finished booting. Let’s start with Vagrant.

Vagrant.configure("2") do |config|
  config.vm.box = "centos/8"

  config.vm.provider :virtualbox do |v|
    v.memory = 4096
  end

  config.vm.provision "shell", path: "user-data.sh"

  config.vm.network "forwarded_port", guest: 80, host: 8080, auto_correct: true
end

To boot this up, once you’ve got Vagrant and Virtualbox installed, run vagrant up and it’ll tell you that it’s set up a port forward from the HTTP port (TCP/80) to a “high” port – TCP/8080. If there’s a collision (because you’re running something else on TCP/8080), it’ll tell you what port it’s forwarded the HTTP port to instead. Once you’ve finished, run vagrant destroy to shut it down. There are lots more tricks you can play with Vagrant, but this is a relatively quick and easy one. Be aware that you’re not using HTTPS, so traffic to the AWX instance can be inspected, but if you’re running this on your local machine, it’s probably not a big issue.

How about running this on a cloud provider, like AWS? We can use the exact same scripts – both the Ansible script, and the user-data.sh script, using Terraform, however, this is a little more complex, as we need to create a VPC, Internet Gateway, Subnet, Security Group and Elastic IP before we can create the virtual machine. What’s more, the Free Tier (that “first hit is free” thing that Amazon Web Services provide to you) does not have enough horsepower to run AWX, so, if you want to look at how to run up AWX in EC2 (or to tweak it to run on Azure, GCP, Digital Ocean or one of the fine offerings from IBM or RedHat), then click through to the gist I’ve put all my code from this post into. The critical lines in there are to select a “CentOS 8” image, open HTTP and SSH into the machine, and to specify the user-data.sh file to provision the machine. Everything else is cruft to make the virtual machine talk to, and be seen by, hosts on the Internet.

To run this one, you need to run terraform init to load the AWS plugin, then terraform apply. Note that this relies on having an AWS access token defined, so if you don’t have them set up, you’ll need to get that sorted out first. Once you’ve finished with your demo, you should run terraform destroy to remove all the assets created by this terraform script. Again, when you’re running that demo, note that you ONLY have HTTP access set up, not HTTPS, so don’t use important credentials on there!

Once you’ve got your AWX environment running, you’ve got just enough AWX there to demo what Ansible Tower looks like, what it can bring to your organisation… and maybe even convince them that it’s worth investing in a license, rather than running AWX in production. Just in case you have that 2AM call-out that we all dread.

Featured image is “pharmacy” by “Tim Evanson” on Flickr and is released under a CC-BY-SA license.

"inventory" by "Lee" on Flickr

Using a AWS Dynamic Inventory with Ansible 2.10

In Ansible 2.10, Ansible started bundling modules and plugins as “Collections”, basically meaning that Ansible didn’t need to make a release every time a vendor wanted to update the libraries it required, or API changes required new fields to be supplied to modules. As part of this split between “Collections” and “Core”, the AWS modules and plugins got moved into a collection.

Now, if you’re using Ansible 2.9 or earlier, this probably doesn’t impact you, but there are some nice features in Ansible 2.10 that I wanted to use, so… buckle up :)

Getting started with Ansible 2.10, using a virtual environment

If you currently are using Ansible 2.9, it’s probably worth creating a “python virtual environment”, or “virtualenv” to try out Ansible 2.10. I did this on my Ubuntu 20.04 machine by typing:

sudo apt install -y virtualenv
mkdir -p ~/bin
cd ~/bin
virtualenv -p python3 ansible_2.10

The above ensures that you have virtualenv installed, creates a directory called “bin” in your home directory, if it doesn’t already exist, and then places the virtual environment, using Python3, into a directory there called “ansible_2.10“.

Whenever we want to use this new environment you must activate it, using this command:

source ~/bin/ansible_2.10/bin/activate

Once you’ve executed this, any binary packages created in that virtual environment will be executed from there, in preference to the file system packages.

You can tell that you’ve “activated” this virtual environment, because your prompt changes from user@HOST:~$ to (ansible_2.10) user@HOST:~$ which helps 😀

Next, let’s create a requirements.txt file. This will let us install the environment in a repeatable manner (which is useful with Ansible). Here’s the content of this file.

ansible>=2.10
boto3
botocore

So, this isn’t just Ansible, it’s also the supporting libraries we’ll need to talk to AWS from Ansible.

We execute the following command:

pip install -r requirements.txt

Note, on Windows Subsystem for Linux version 1 (which I’m using) this will take a reasonable while, particularly if it’s crossing from the WSL environment into the Windows environment, depending on where you have specified the virtual environment to be placed.

If you get an error message about something to do with being unable to install ffi, then you’ll need to install the package libffi-dev with sudo apt install -y libffi-dev and then re-run the pip install command above.

Once the installation has completed, you can run ansible --version to see something like the following:

ansible 2.10.2
  config file = None
  configured module search path = ['/home/user/.ansible/plugins/modules', '/usr/share/ansible/plugins/modules']
  ansible python module location = /home/user/ansible_2.10/lib/python3.8/site-packages/ansible
  executable location = /home/user/ansible_2.10/bin/ansible
  python version = 3.8.2 (default, Jul 16 2020, 14:00:26) [GCC 9.3.0]

Configuring Ansible for local collections

Ansible relies on certain paths in the filesystem to store things like collections, roles and modules, but I like to circumvent these things – particularly if I’m developing something, or moving from one release to the next. Fortunately, Ansible makes this very easy, using a single file, ansible.cfg to tell the code that’s running in this path where to find things.

A quick note on File permissions with ansible.cfg

Note that the POSIX file permissions for the directory you’re in really matter! It must be set to 775 (-rwxrwxr-x) as a maximum – if it’s “world writable” (the last number) it won’t use this file! Other options include 770, 755. If you accidentally set this as world writable, or are using a directory from the “Windows” side of WSL, then you’ll get an error message like this:

[WARNING]: Ansible is being run in a world writable directory (/home/user/ansible_2.10_aws), ignoring it as an ansible.cfg source. For more information see
https://docs.ansible.com/ansible/devel/reference_appendices/config.html#cfg-in-world-writable-dir

That link is this one: https://docs.ansible.com/ansible/devel/reference_appendices/config.html#cfg-in-world-writable-dir and has some useful advice.

Back to configuring Ansible

In ansible.cfg, I have the following configured:

[defaults]
collections_paths = ./collections:~/.ansible/collections:/usr/share/ansible/collections

This file didn’t previously exist in this directory, so I created that file.

This block asks Ansible to check the following paths in order:

  • collections in this path (e.g. /home/user/ansible_2.10_aws/collections)
  • collections in the .ansible directory under the user’s home directory (e.g. /home/user/.ansible/collections)
  • and finally /usr/share/ansible/collections for system-wide collections.

If you don’t configure Ansible with the ansible.cfg file, the default is to store the collections in ~/.ansible/collections, but you can “only have one version of the collection”, so this means that if you’re relying on things not to change when testing, or if you’re running multiple versions of Ansible on your system, then it’s safest to store the collections in the same file tree as you’re working in!

Installing Collections

Now we have Ansible 2.10 installed, and our Ansible configuration file set up, let’s get our collection ready to install. We do this with a requirements.yml file, like this:

---
collections:
- name: amazon.aws
  version: ">=1.2.1"

What does this tell us? Firstly, that we want to install the Amazon AWS collection from Ansible Galaxy. Secondly that we want at least the most current version (which is currently version 1.2.1). If you leave the version line out, it’ll get “the latest” version. If you replace ">=1.2.1" with 1.2.1 it’ll install exactly that version from Galaxy.

If you want any other collections, you add them as subsequent lines (more details here), like this:

collections:
- name: amazon.aws
  version: ">=1.2.1"
- name: some.other
- name: git+https://example.com/someorg/somerepo.git
  version: 1.0.0
- name: git@example.com:someorg/someotherrepo.git

Once we’ve got this file, we run this command to install the content of the requirements.yml: ansible-galaxy collection install -r requirements.yml

In our case, this installs just the amazon.aws collection, which is what we want. Fab!

Getting our dynamic inventory

Right, so we’ve got all the pieces now that we need! Let’s tell Ansible that we want it to ask AWS for an inventory. There are three sections to this.

Configuring Ansible, again!

We need to open up our ansible.cfg file. Because we’re using the collection to get our Dynamic Inventory plugin, we need to tell Ansible to use that plugin. Edit ./ansible.cfg in your favourite editor, and add this block to the end:

[inventory]
enable_plugins = aws_ec2

If you previously created the ansible.cfg file when you were setting up to get the collection installed alongside, then your ansible.cfg file will look (something) like this:

[defaults]
collections_paths     = ./collections:~/.ansible/collections:/usr/share/ansible/collections

[inventory]
enable_plugins = amazon.aws.aws_ec2

Configure AWS

Your machine needs to have access tokens to interact with the AWS API. These are stored in ~/.aws/credentials (e.g. /home/user/.aws/credentials) and look a bit like this:

[default]
aws_access_key_id = A1B2C3D4E5F6G7H8I9J0
aws_secret_access_key = A1B2C3D4E5F6G7H8I9J0a1b2c3d4e5f6g7h8i9j0

Set up your inventory

In a bit of a change to how Ansible usually does the inventory, to have a plugin based dynamic inventory, you can’t specify a file any more, you have to specify a directory. So, create the file ./inventory/aws_ec2.yaml (having created the directory inventory first). The file contains the following:

---
plugin: amazon.aws.aws_ec2

Late edit 2020-12-01: Further to the comment by Giovanni, I’ve amended this file snippet from plugin: aws_ec2 to plugin: amazon.aws.aws_ec2.

By default, this just retrieves the hostnames of any running EC2 instance, as you can see by running ansible-inventory -i inventory --graph

@all:
  |--@aws_ec2:
  |  |--ec2-176-34-76-187.eu-west-1.compute.amazonaws.com
  |  |--ec2-54-170-131-24.eu-west-1.compute.amazonaws.com
  |  |--ec2-54-216-87-131.eu-west-1.compute.amazonaws.com
  |--@ungrouped:

I need a bit more detail than this – I like to use the tags I assign to AWS assets to decide what I’m going to target the machines with. I also know exactly which regions I’ve got my assets in, and what I want to use to get the names of the devices, so this is what I’ve put in my aws_ec2.yaml file:

---
plugin: amazon.aws.aws_ec2
keyed_groups:
- key: tags
  prefix: tag
- key: 'security_groups|json_query("[].group_name")'
  prefix: security_group
- key: placement.region
  prefix: aws_region
- key: tags.Role
  prefix: role
regions:
- eu-west-1
hostnames:
- tag:Name
- dns-name
- public-ip-address
- private-ip-address

Late edit 2020-12-01: Again, I’ve amended this file snippet from plugin: aws_ec2 to plugin: amazon.aws.aws_ec2.

Now, when I run ansible-inventory -i inventory --graph, I get this output:

@all:
  |--@aws_ec2:
  |  |--euwest1-firewall
  |  |--euwest1-demo
  |  |--euwest1-manager
  |--@aws_region_eu_west_1:
  |  |--euwest1-firewall
  |  |--euwest1-demo
  |  |--euwest1-manager
  |--@role_Firewall:
  |  |--euwest1-firewall
  |--@role_Firewall_Manager:
  |  |--euwest1-manager
  |--@role_VM:
  |  |--euwest1-demo
  |--@security_group_euwest1_allow_all:
  |  |--euwest1-firewall
  |  |--euwest1-demo
  |  |--euwest1-manager
  |--@tag_Name_euwest1_firewall:
  |  |--euwest1-firewall
  |--@tag_Name_euwest1_demo:
  |  |--euwest1-demo
  |--@tag_Name_euwest1_manager:
  |  |--euwest1-manager
  |--@tag_Role_Firewall:
  |  |--euwest1-firewall
  |--@tag_Role_Firewall_Manager:
  |  |--euwest1-manager
  |--@tag_Role_VM:
  |  |--euwest1-demo
  |--@ungrouped:

To finish

Now you have your dynamic inventory, you can target your playbook at any of the groups listed above (like role_Firewall, aws_ec2, aws_region_eu_west_1 or some other tag) like you would any other inventory assignment, like this:

---
- hosts: role_Firewall
  gather_facts: false
  tasks:
  - name: Show the name of this device
    debug:
      msg: "{{ inventory_hostname }}"

And there you have it. Hope this is useful!

Late edit: 2020-11-23: Following a conversation with Andy from Work, we’ve noticed that if you’re trying to do SSM connections, rather than username/password based ones, you might want to put this in your aws_ec2.yml file:

---
plugin: amazon.aws.aws_ec2
hostnames:
  - tag:Name
compose:
  ansible_host: instance_id
  ansible_connection: 'community.aws.aws_ssm'

Late edit 2020-12-01: One final instance, I’ve changed plugin: aws_ec2 to plugin: amazon.aws.aws_ec2.

This will keep your hostnames “pretty” (with whatever you’ve tagged it as), but will let you connect over SSM to the Instance ID. Good fun :)

Featured image is “inventory” by “Lee” on Flickr and is released under a CC-BY-SA license.

"centos login" by "fsse8info" on Flickr

Getting the default username and AMI for an OS with #Terraform

I have a collection of AWS AMIs I use for various builds at work. These come from two places – the AWS Marketplace and our internal Build process.

Essentially, our internal builds (for those who work for my employer – these are the OptiMISe builds) are taken from specific AWS Marketplace builds and hardened.

Because I don’t want to share the AMI details when I put stuff on GitHub, I have an override.tf file that handles the different AMI search strings. So, here’s the ami.tf file I have with the AWS Marketplace version:

data "aws_ami" "centos7" {
  most_recent = true

  filter {
    name   = "name"
    values = ["CentOS Linux 7 x86_64 HVM EBS ENA*"]
  }

  filter {
    name   = "architecture"
    values = ["x86_64"]
  }

  owners = ["679593333241"] # CentOS Project
}

And here’s an example of the override.tf file I have:

data "aws_ami" "centos7" {
  most_recent = true

  filter {
    name   = "name"
    values = ["SomeUniqueString Containing CentOS*"]
  }

  owners = ["123456789012"]
}

Next, I put these AMI images into a “null” data source, which is evaluated at runtime:

data "null_data_source" "os" {
  inputs = {
    centos7 = data.aws_ami.centos7.id
  }
}

I always forget which username goes with each image, so in the ami.tf file, I also have this:

variable "username" {
  type = map(string)
  default = {
    centos7 = "centos"
  }
}

And in the override.tf file, I have this:

variable "username" {
  type = map(string)
  default = {
    centos7 = "someuser"
  }
}

To get the right combination of username and AMI, I have this in the file where I create my “instance” (virtual machine):

variable "os" {
  default = "centos7"
}

resource "aws_instance" "vm01" {
  ami = data.null_data_source.os.outputs[var.os]
  # additional lines omitted for brevity
}

output "username" {
  value = var.username[var.os]
}

output "vm01" {
  value = aws_instance.vm01.public_ip
}

And that way, I get the VM’s default username and IP address on build. Nice.

Late edit – 2020-09-20: It’s worth noting that this is fine for short-lived builds, proof of concept, etc. But, for longer lived environments, you should be calling out exactly which AMI you’re using, right from the outset. That way, your builds will (or should) all start out from the same point, no ambiguity about exactly which point release they’re getting, etc.

Featured image is “centos login” by “fsse8info” on Flickr and is released under a CC-BY-SA license.

"Tracking Methane Sources and Movement Around the Globe" by "NASA/Scientific Visualization Studio" on Nasa.gov

Flexibly loading files in Terraform to license a FortiGate firewall on AWS, Azure and other Cloud platforms

One of the things I’m currently playing with is a project to deploy some FortiGate Firewalls into cloud platforms. I have a couple of Evaluation Licenses I can use (as we’re a partner), but when it comes to automatically scaling, you need to use the PAYG license.

To try to keep my terraform files as reusable as possible, I came up with this work around. It’s likely to be useful in other places too. Enjoy!

This next block is stored in license.tf and basically says “by default, you have no license.”

variable "license_file" {
  default = ""
  description = "Path to the license file to load, or leave blank to use a PAYG license."
}

We can either override this with a command line switch terraform apply -var 'license_file=mylicense.lic', or (more likely) the above override file named license_override.tf (ignored in Git) which has this next block in it:

variable "license_file" {
  default = "mylicense.lic"
}

This next block is also stored in license.tf and says “If var.license is not empty, load that license file [var.license != "" ? var.license] but if it is empty, check whether /dev/null exists (*nix platforms) [fileexists("/dev/null")] in which case, use /dev/null, otherwise use the NUL: device (Windows platforms).”

data "local_file" "license" {
  filename = var.license_file != "" ? var.license_file : fileexists("/dev/null") ? "/dev/null" : "NUL:"
}

👉 Just as an aside, I’ve seen this “ternary” construct in a few languages. It basically looks like this: boolean_operation ? true_value : false_value

That check, logically, could have been written like this instead: "%{if boolean_operation}${true_value}%{else}${false_value}%{endif}"

By combining two of these together, while initially it looks far more messy and hard to parse, I’ve found that, especially in single-line statements, it’s much more compact and eventually easier to read than the alternative if/else/endif structure.

So, this means that we can now refer to data.local_file.license as our data source.

Next, I want to select either the PAYG (Pay As You Go) or BYOL (Bring Your Own License) licensed AMI in AWS (the same principle applies in Azure, GCP, etc), so in this block we provide a different value to the filter in the AMI Data Source, suggesting the string “FortiGate-VM64-AWS *x.y.z*” if we have a value provided license, or “FortiGate-VM64-AWSONDEMAND *x.y.z*” if we don’t.

data "aws_ami" "FortiGate" {
  most_recent = true

  filter {
    name   = "name"
    values = ["FortiGate-VM64-AWS%{if data.local_file.license.content == ""}ONDEMAND%{endif} *${var.release}*"]
  }

  filter {
    name   = "virtualization-type"
    values = ["hvm"]
  }

  owners = ["679593333241"] # AWS
}

And the very last thing is to create the user-data template (known as customdata in Azure), using this block:

data "template_cloudinit_config" "config" {
  gzip          = false
  base64_encode = false

  part {
    filename     = "config"
    content_type = "multipart/mixed"
    content      = templatefile(
      "${path.module}/user_data.txt.tmpl",
      {
        hostname = "firewall"
      }
    )
  }

  part {
    filename     = "license"
    content_type = "text/plain"
    content      = data.local_file.license.content
  }
}

And so that is how I can elect to provide a license, or use a pre-licensed image from AWS, and these lessons can also be applied in an Azure or GCP environment too.

Featured image is “Tracking Methane Sources and Movement Around the Globe” by “NASA/Scientific Visualization Studio

"Field Notes - Sweet Tooth" by "The Marmot" on Flickr

Multi-OS builds in AWS with Terraform – some notes from the field!

Late edit: 2020-05-22 – Updated with better search criteria from colleague conversations

I’m building a proof of concept for … well, a product that needs testing on several different Linux and Windows variants on AWS and Azure. I’m building this environment with Terraform, and it’s thrown me a few curve balls, so I thought I’d document the issues I’ve had!

The versions of distributions I have tested are the latest releases of each of these images at-or-near the time of writing. The major version listed is the earliest I have tested, so no assumption is made about previous versions, and later versions, after the time of this post should not assume any of this data is also accurate!

(Fujitsu Staff – please contact me on my work email address for details on how to get the internal AMIs of our builds of these images 😄)

Linux Distributions

On the whole, I tend to be much more confident and knowledgable about Linux distributions. I’ve also done far more installs of each of these!

Almost all of these installs are Free of Charge, with the exception of Red Hat Enterprise Linux, which requires a subscription fee, and this can be “Pay As You Go” or “Bring Your Own License”. These sorts of things are arranged for me, so I don’t know how easy or hard it is to organise these licenses!

These builds all use cloud-init, via either a cloud-init yaml script, or some shell scripting language (usually accepted to be bash). If this script fails to execute, you will find your user-data file in /var/lib/cloud/instance/scripts/part-001. If this is a shell script then you will be able to execute it by running that script as your root user.

Amazon Linux 2 or Amzn2

Amazon Linux2 is the “preferred” distribution for Amazon Web Services (AWS) (surprisingly enough). It is based on Red Hat Enterprise Linux (RHEL), and many of the instructions you’ll want to run to install software will use RHEL based instructions. This platform is not available outside the AWS ecosystem, as far as I can tell, although you might be able to run it on-prem.

Software packages are limited in this distribution, so any “extra” features require the installation of the “EPEL” repository, by executing the command sudo amazon-linux-extras install epel and then using the yum command to install further packages. I needed nginx for part of my build, and this was only in EPEL.

Amzn2 AMI Lookup

data "aws_ami" "amzn2" {
  most_recent = true

  filter {
    name   = "name"
    values = ["amzn2-ami-hvm-2.0.*-gp2"]
  }

  filter {
    name   = "architecture"
    values = ["x86_64"]
  }

  filter {
    name   = "state"
    values = ["available"]
  }

  owners = ["amazon"] # Canonical
}

Amzn2 User Account

Amazon Linux 2 images under AWS have a default “ec2-user” user account. sudo will allow escalation to Root with no password prompt.

Amzn2 AWS Interface Configuration

The primary interface is called eth0. Network Manager is not installed. To manage the interface, you need to edit /etc/sysconfig/network-scripts/ifcfg-eth0 and apply changes with ifdown eth0 ; ifup eth0.

Amzn2 user-data / Cloud-Init Troubleshooting

I’ve found the output from user-data scripts appearing in /var/log/cloud-init-output.log.

CentOS 7

For starters, AWS doesn’t have an official CentOS8 image, so I’m a bit stymied there! In fact, as far as I can make out, CentOS is only releasing ISOs for builds now, and not any cloud images. There’s an open issue on their bug tracker which seems to suggest that it’s not going to get any priority any time soon! Blimey.

This image may require you to “subscribe” to the image (particularly if you have a “private marketplace”), but this will be requested of you (via a URL provided on screen) when you provision your first machine with this AMI.

Like with Amzn2, CentOS7 does not have nginx installed, and like Amzn2, installation of the EPEL library is not a difficult task. CentOS7 bundles a file to install the EPEL, installed by running yum install epel-release. After this is installed, you have the “full” range of software in EPEL available to you.

CentOS AMI Lookup

data "aws_ami" "centos7" {
  most_recent = true

  filter {
    name   = "name"
    values = ["CentOS Linux 7*"]
  }

  filter {
    name   = "architecture"
    values = ["x86_64"]
  }

  filter {
    name   = "state"
    values = ["available"]
  }

  owners = ["aws-marketplace"]
}

CentOS User Account

CentOS7 images under AWS have a default “centos” user account. sudo will allow escalation to Root with no password prompt.

CentOS AWS Interface Configuration

The primary interface is called eth0. Network Manager is not installed. To manage the interface, you need to edit /etc/sysconfig/network-scripts/ifcfg-eth0 and apply changes with ifdown eth0 ; ifup eth0.

CentOS Cloud-Init Troubleshooting

I’ve run several different user-data located bash scripts against this system, and the logs from these scripts are appearing in the default syslog file (/var/log/syslog) or by running journalctl -xefu cloud-init. They do not appear in /var/log/cloud-init-output.log.

Red Hat Enterprise Linux (RHEL) 7 and 8

Red Hat has both RHEL7 and RHEL8 images in the AWS market place. The Proof Of Value (POV) I was building was only looking at RHEL7, so I didn’t extensively test RHEL8.

Like Amzn2 and CentOS7, RHEL7 needs EPEL installing to have additional packages installed. Unlike Amzn2 and CentOS7, you need to obtain the EPEL package from the Fedora Project. Do this by executing these two commands:

wget https://dl.fedoraproject.org/pub/epel/epel-release-latest-7.noarch.rpm
yum install epel-release-latest-7.noarch.rpm

After this is installed, you’ll have access to the broader range of software that you’re likely to require. Again, I needed nginx, and this was not available to me with the stock install.

RHEL7 AMI Lookup

data "aws_ami" "rhel7" {
  most_recent = true

  filter {
    name   = "name"
    values = ["RHEL-7*GA*Hourly*"]
  }

  filter {
    name   = "architecture"
    values = ["x86_64"]
  }

  filter {
    name   = "state"
    values = ["available"]
  }

  owners = ["309956199498"] # Red Hat
}

RHEL8 AMI Lookup

data "aws_ami" "rhel8" {
  most_recent = true

  filter {
    name   = "name"
    values = ["RHEL-8*HVM-*Hourly*"]
  }

  filter {
    name   = "architecture"
    values = ["x86_64"]
  }

  filter {
    name   = "state"
    values = ["available"]
  }

  owners = ["309956199498"] # Red Hat
}

RHEL User Accounts

RHEL7 and RHEL8 images under AWS have a default “ec2-user” user account. sudo will allow escalation to Root with no password prompt.

RHEL AWS Interface Configuration

The primary interface is called eth0. Network Manager is installed, and the eth0 interface has a profile called “System eth0” associated to it.

RHEL Cloud-Init Troubleshooting

In RHEL7, as per CentOS7, logs from user-data scripts are appear in the general syslog file (in this case, /var/log/messages) or by running journalctl -xefu cloud-init. They do not appear in /var/log/cloud-init-output.log.

In RHEL8, logs from user-data scrips now appear in /var/log/cloud-init-output.log.

Ubuntu 18.04

At the time of writing this, the vendor, who’s product I was testing, categorically stated that the newest Ubuntu LTS, Ubuntu 20.04 (Focal Fossa) would not be supported until some time after our testing was complete. As such, I spent no time at all researching or planning to use this image.

Ubuntu is the only non-RPM based distribution in this test, instead being based on the Debian project’s DEB packages. As such, it’s range of packages is much wider. That said, for the project I was working on, I required a later version of nginx than was available in the Ubuntu Repositories, so I had to use the nginx Personal Package Archive (PPA). To do this, I found the official PPA for the nginx project, and followed the instructions there. Generally speaking, this would potentially risk any support from the distribution vendor, as it’s not certified or supported by the project… but I needed that version, so I had to do it!

Ubuntu 18.04 AMI Lookup

data "aws_ami" "ubuntu1804" {
  most_recent = true

  filter {
    name   = "name"
    values = ["*ubuntu*18.04*"]
  }

  filter {
    name   = "architecture"
    values = ["x86_64"]
  }

  filter {
    name   = "state"
    values = ["available"]
  }

  owners = ["099720109477"] # Canonical
}

Ubuntu 18.04 User Accounts

Ubuntu 18.04 images under AWS have a default “ubuntu” user account. sudo will allow escalation to Root with no password prompt.

Ubuntu 18.04 AWS Interface Configuration

The primary interface is called eth0. Network Manager is not installed, and instead Ubuntu uses Netplan to manage interfaces. The file to manage the interface defaults is /etc/netplan/50-cloud-init.yaml. If you struggle with this method, you may wish to install ifupdown and define your configuration in /etc/network/interfaces.

Ubuntu 18.04 Cloud-Init Troubleshooting

In Ubuntu 18.04, logs from user-data scrips appear in /var/log/cloud-init-output.log.

Windows

This section is far more likely to have it’s data consolidated here!

Windows has a common “standard” username – Administrator, and a common way of creating a password (this is generated on-boot, and the password is transferred to the AWS Metadata Service, which it is retrieved and decrypted with the SSH key you’ve used to build the “authentication” to the box) which Terraform handles quite nicely.

The network device is referred to as “AWS PV Network Device #0”. It can be managed with powershell, netsh (although apparently Microsoft are rumbling about demising this script), or from the GUI.

Windows 2012R2

This version is very old now, and should be compared to Windows 7 in terms of age. It is only supported by Microsoft with an extended maintenance package!

Windows 2012R2 AMI Lookup

data "aws_ami" "w2012r2" {
  most_recent = true

  filter {
    name = "name"
    values = ["Windows_Server-2012-R2_RTM-English-64Bit-Base*"]
  }

  filter {
    name   = "architecture"
    values = ["x86_64"]
  }

  filter {
    name   = "state"
    values = ["available"]
  }

  owners = ["801119661308"] # AWS
}

Windows 2012R2 Cloud-Init Troubleshooting

Logs from the Metadata Service can be found in C:\Program Files\Amazon\Ec2ConfigService\Logs\Ec2ConfigLog.txt. You can also find the userdata script in C:\Program Files\Amazon\Ec2ConfigService\Scripts\UserScript.ps1. This can be launched and debugged using PowerShell ISE, which is in the “Start” menu.

Windows 2016

This version is reasonably old now, and should be compared to Windows 8 in terms of age. It is supported until 2022 in “mainline” support.

Windows 2016 AMI Lookup

data "aws_ami" "w2016" {
  most_recent = true

  filter {
    name = "name"
    values = ["Windows_Server-2016-English-Full-Base*"]
  }

  filter {
    name   = "architecture"
    values = ["x86_64"]
  }

  filter {
    name   = "state"
    values = ["available"]
  }

  owners = ["801119661308"] # AWS
}

Windows 2016 Cloud-Init Troubleshooting

The metadata service has moved from Windows 2016 and onwards. Logs are stored in a partially hidden directory tree, so you may need to click in the “Address” bar of the Explorer window and type in part of this path. The path to these files is: C:\ProgramData\Amazon\EC2-Windows\Launch\Log. I say “files” as there are two parts to this file – an “Ec2Launch.log” file which reports on the boot process, and “UserdataExecution.log” which shows the output from the userdata script.

Unlike with the Windows 2012R2 version, you can’t get hold of the actual userdata script on the filesystem, you need to browse to a special path in the metadata service (actually, technically, you can do this with any of the metadata services – OpenStack, Azure, and so on) which is: http://169.254.169.254/latest/user-data/

This will contain userdata between a <powershell> and </powershell> pair of tags. This would need to be copied out of this URL and pasted into a new file on your local machine to determine why issues are occurring. Again, I would recommend using PowerShell ISE from the Start Menu to debug your code.

Windows 2019

This version is the most recent released version of Windows Server, and should be compared to Windows 10 in terms of age.

Windows 2019 AMI Lookup

data "aws_ami" "w2019" {
  most_recent = true

  filter {
    name = "name"
    values = ["Windows_Server-2019-English-Full-Base*"]
  }

  filter {
    name   = "architecture"
    values = ["x86_64"]
  }

  filter {
    name   = "state"
    values = ["available"]
  }

  owners = ["801119661308"] # AWS
}

Windows 2019 Cloud-Init Troubleshooting

Functionally, the same as Windows 2016, but to recap, the metadata service has moved from Windows 2016 and onwards. Logs are stored in a partially hidden directory tree, so you may need to click in the “Address” bar of the Explorer window and type in part of this path. The path to these files is: C:\ProgramData\Amazon\EC2-Windows\Launch\Log. I say “files” as there are two parts to this file – an “Ec2Launch.log” file which reports on the boot process, and “UserdataExecution.log” which shows the output from the userdata script.

Unlike with the Windows 2012R2 version, you can’t get hold of the actual userdata script on the filesystem, you need to browse to a special path in the metadata service (actually, technically, you can do this with any of the metadata services – OpenStack, Azure, and so on) which is: http://169.254.169.254/latest/user-data/

This will contain userdata between a <powershell> and </powershell> pair of tags. This would need to be copied out of this URL and pasted into a new file on your local machine to determine why issues are occurring. Again, I would recommend using PowerShell ISE from the Start Menu to debug your code.

Featured image is “Field Notes – Sweet Tooth” by “The Marmot” on Flickr and is released under a CC-BY license.

“New shoes” by “Morgaine” from Flickr

Making Windows Cloud-Init Scripts run after a reboot (Using Terraform)

I’m currently building a Proof Of Value (POV) environment for a product, and one of the things I needed in my environment was an Active Directory domain.

To do this in AWS, I had to do the following steps:

  1. Build my Domain Controller
    1. Install Windows
    2. Set the hostname (Reboot)
    3. Promote the machine to being a Domain Controller (Reboot)
    4. Create a domain user
  2. Build my Member Server
    1. Install Windows
    2. Set the hostname (Reboot)
    3. Set the DNS client to point to the Domain Controller
    4. Join the server to the domain (Reboot)

To make this work, I had to find a way to trigger build steps after each reboot. I was working with Windows 2012R2, Windows 2016 and Windows 2019, so the solution had to be cross-version. Fortunately I found this script online! That version was great for Windows 2012R2, but didn’t cover Windows 2016 or later… So let’s break down what I’ve done!

In your userdata field, you need to have two sets of XML strings, as follows:

<persist>true</persist>
<powershell>
$some = "powershell code"
</powershell>

The first block says to Windows 2016+ “keep trying to run this script on each boot” (note that you need to stop it from doing non-relevant stuff on each boot – we’ll get to that in a second!), and the second bit is the PowerShell commands you want it to run. The rest of this now will focus just on the PowerShell block.

  $path= 'HKLM:\Software\UserData'
  
  if(!(Get-Item $Path -ErrorAction SilentlyContinue)) {
    New-Item $Path
    New-ItemProperty -Path $Path -Name RunCount -Value 0 -PropertyType dword
  }
  
  $runCount = Get-ItemProperty -Path $path -Name Runcount -ErrorAction SilentlyContinue | Select-Object -ExpandProperty RunCount
  
  if($runCount -ge 0) {
    switch($runCount) {
      0 {
        $runCount = 1 + [int]$runCount
        Set-ItemProperty -Path $Path -Name RunCount -Value $runCount
        if ($ver -match 2012) {
          #Enable user data
          $EC2SettingsFile = "$env:ProgramFiles\Amazon\Ec2ConfigService\Settings\Config.xml"
          $xml = [xml](Get-Content $EC2SettingsFile)
          $xmlElement = $xml.get_DocumentElement()
          $xmlElementToModify = $xmlElement.Plugins
          
          foreach ($element in $xmlElementToModify.Plugin)
          {
            if ($element.name -eq "Ec2HandleUserData") {
              $element.State="Enabled"
            }
          }
          $xml.Save($EC2SettingsFile)
        }
        $some = "PowerShell Script"
      }
    }
  }

Whew, what a block! Well, again, we can split this up into a couple of bits.

In the first few lines, we build a pointer, a note which says “We got up to here on our previous boots”. We then read that into a variable and find that number and execute any steps in the block with that number. That’s this block:

  $path= 'HKLM:\Software\UserData'
  
  if(!(Get-Item $Path -ErrorAction SilentlyContinue)) {
    New-Item $Path
    New-ItemProperty -Path $Path -Name RunCount -Value 0 -PropertyType dword
  }
  
  $runCount = Get-ItemProperty -Path $path -Name Runcount -ErrorAction SilentlyContinue | Select-Object -ExpandProperty RunCount
  
  if($runCount -ge 0) {
    switch($runCount) {

    }
  }

The next part (and you’ll repeat it for each “number” of reboot steps you need to perform) says “increment the number” then “If this is Windows 2012, remind the userdata handler that the script needs to be run again next boot”. That’s this block:

      0 {
        $runCount = 1 + [int]$runCount
        Set-ItemProperty -Path $Path -Name RunCount -Value $runCount
        if ($ver -match 2012) {
          #Enable user data
          $EC2SettingsFile = "$env:ProgramFiles\Amazon\Ec2ConfigService\Settings\Config.xml"
          $xml = [xml](Get-Content $EC2SettingsFile)
          $xmlElement = $xml.get_DocumentElement()
          $xmlElementToModify = $xmlElement.Plugins
          
          foreach ($element in $xmlElementToModify.Plugin)
          {
            if ($element.name -eq "Ec2HandleUserData") {
              $element.State="Enabled"
            }
          }
          $xml.Save($EC2SettingsFile)
        }
        
      }

In fact, it’s fair to say that in my userdata script, this looks like this:

  $path= 'HKLM:\Software\UserData'
  
  if(!(Get-Item $Path -ErrorAction SilentlyContinue)) {
    New-Item $Path
    New-ItemProperty -Path $Path -Name RunCount -Value 0 -PropertyType dword
  }
  
  $runCount = Get-ItemProperty -Path $path -Name Runcount -ErrorAction SilentlyContinue | Select-Object -ExpandProperty RunCount
  
  if($runCount -ge 0) {
    switch($runCount) {
      0 {
        ${file("templates/step.tmpl")}

        ${templatefile(
          "templates/rename_windows.tmpl",
          {
            hostname = "SomeMachine"
          }
        )}
      }
      1 {
        ${file("templates/step.tmpl")}

        ${templatefile(
          "templates/join_ad.tmpl",
          {
            dns_ipv4 = "192.0.2.1",
            domain_suffix = "ad.mycorp",
            join_account = "ad\someuser",
            join_password = "SomePassw0rd!"
          }
        )}
      }
    }
  }

Then, after each reboot, you need a new block. I have a block to change the computer name, a block to join the machine to the domain, and a block to install an software that I need.

Featured image is “New shoes” by “Morgaine” on Flickr and is released under a CC-BY-SA license.