cloud-foundation-fabric/infrastructure/onprem-google-access-dns
Ludovico Magnocavallo c486bfc66f
Merge development branch (#44)
* VPN-HA module initial commit

* Added readme for net-vpn-ha module

* Update readme, add simple description

* Merge new modules list and environments foundation example (#30)

* gke-cluster

* net-vpc module and tests

* add TODO to net-vpc module

* add minimal README files with input/output variables to gke and net-vpc modules

* BigQuery Module (#24)

* Bigquery Module

* Added README file

* Added type hints

* gke-cluster

* net-vpc module and tests

* add TODO to net-vpc module

* add minimal README files with input/output variables to gke and net-vpc modules

* BigQuery Module (#24)

* Bigquery Module

* Added README file

* Added type hints

* GCS module

* net vpc module: improve secondary range outputs

* net vpc module: add serve project registration

* project module

* move bigquery module to not-ready folder

* folders module

* rename project module's iam variables

* slight tweak to folder module outputs

* gcs module

* simplify net-vpc module variables

* fix module tests configurations, fix net-vpc module tests

* add pydoc utility

* add/update module READMEs

* add/update module READMEs

* add/update module READMEs

* improve variable type summary generation in tfdoc

* tfdoc: add support for replacing doc in README.md files

* improve module READMEs

* net-vpc-firewall module

* add support for sensitive output attribute in tfdoc

* remove empty function from tfdoc

* render variable type as code in tfdoc

* update module READMEs

* net address module

* net cloudnat module

* remove redundant variable from net-cloudnat module

* vpc module: add support for peering, use network name as subnet name prefix

* net-vpn-static module

* net-vpn-static module README

* net-vpn-static module README

* tfdoc: fix error on undeclared variable type

* dns module

* set version for all modules

* kms module (untested)

* change kms key self links output to map, fix gcs and kms iam variable descriptions

* fix kms module

* update kms module readme

* simplify local iam pairs in modules

* service accounts module (unfinished)

* work on service accounts module

* project module: add gcr service account

* project module: update outputs in README

* first working version of the iam service accounts module

* iam service accounts module: extra checks in locals

* modules/net-cloudnat: reorder variables

* modules/net-vpn-dynamic: initial import (untested)

* modules/net-vpn-dynamic: first working version

* modules/net-vpn-dynamic: add outputs for auto-created router

* modules/net-vpn-dynamic: update README

* modules/net-[vpn,cloudnat]: clean up variable,s remove prefix

* modules/net-vpn-dynamic: add advertisement configuration to tunnel bgp peer, refactor variables

* tfdoc: add tooltips for variable types and defaults

* modules: update README variables and outputs

* tfdoc: improve variable default rendering

* modules: update README variables and outputs

* modules/net-vpc: minimal output refactoring

* modules/vm-cos: initial import, base resources working, no outputs

* modules/vm-cos: add variable descriptions

* tfdoc: fix parsing in type and default blocks

* modules/vm-cos: fix README

* tfdoc: fix parsing in type and default blocks

* modules/vm-cos: fix README

* modules/compute-vm: initial working import (not fully tested)

* modules/vm-cos: move to not-ready

* tfdoc: fix variable defaults formatting

* modules: update README files with tfdoc fixes

* modules: add initial examples

* gke-nodepool: initial import, untested

* gke nodepool: add README, fix location variable, set node count default to 1

* gke cluster: fix private cluster variables

* gke nodepool: fix README title

* gke cluster: add output for cluster location

* gke nodepool: add missing variables for project id and cluster name, remove default from location variable, fix gke version assignment

* gke nodepool: update README

* net-cloudnat: fix router name when creating default router

* fix variables used for address and router optional creation

* vpn dynamic: fix README

* modules/net-vpn-dynamic: fix router name output

* modules/compute-vm: remove unused variable

* modules/compute-vm-cos-coredns: initial import

* Update foundations modules versions (#26)

* update foundations modules versions

* update Terraform version to v0.12.19 in CI test configuration

* backport tfdoc from Ludo's branch (#27)

* Update docs using tfdoc format (#28)

* update README files

* set all types on variables

* foundations/environments: move log filter to a variable, use org for xpn by default

* foundations/environments: do not use liens by default

* modules/ntp-vpc: better shared_vpc_host variable description

* modules/logging-sinks: initial version

* modules/logging-sinks: streamline options in sinks variable

* modules/compute-vm-cos-coredns: add support for additional files

* modules/folders: rename from 'folder'

* modules/logging-sinks: fix circular dependencies and improve variables

* modules/project: remove extra variable

* modules/bigquery: new module with dataset support only

* foundations/environments: refactor using local modules

* modules/bigquery: better variables, README description and example

* modules: fix a few READMEs

Co-authored-by: Julio Castillo <juliocc@gmail.com>

* modules/net-vpc: README description and examples

* modules/net-vpc: tweak README description and examples

* modules/net-vpc: tweak README description and examples

* modules/net-vpc-firewall: change tag-based rule default ranges, improve README examples and description

* modules/compute-vm: README changes

* modules/compute-vm: use an object for the service account variable, update README

* modules/compute-vm: update README variables table

* modules/compute-vm: add TODO list to README

* modules/compute-vm: add TODO list to README

* modules/compute-vm: add outputs for service account

* modules/net-cloudnat: README

* modules/net-cloudnat: README

* modules/net-cloudnat: add router_create variable

* modules/compute-vm: simplify service account variables

* modules/net-vpn-dynamic: fix README example, use local secret for both empty string and null

* modules/net-vpn-dynamic: improve README example

* modules/gke-cluster: minimal README tweaks

* modules/kms: fix ephemeral keys resource name

* modules/iam-service-accounts: add storage roles

* modules/gke-nodepool: fix node default scopes

* New project variable to prevent deletion of default network (#32)

* New project variable to prevent deletion of default network

This is a workaround to fix
terraform-google-modules/cloud-foundation-fabric#31 while the GCP
terraform provider is fixed

* Add TODOs to remove workarounds in the project module

* Fix Cloud Build files

* modules/gke-nodepool: add monitoring scope to defaults

* modules/iam-service-accounts: add support for IAM bindings onthe service accounts

* playground module in sandbox, remove not ready modules

* Fix ci configurations in development branch (#33)

* try fixing ci confgurations

* add exclusion match to ci boilerplate check

* add skip boilerplate comment to compute-vm-cos-coredns template fragment

* modules/gke-cluster: fix boilerplate in outputs

* Simplify tests, re-enable CI

* add instance group support to compute-vm, start tests refactoring

* modules/compute-vm: group fixes, tests

* modules/compute-vm: minimal test beautification

* simplify top-level pytest fixture

* modules/dns: tests and minor tweaks

* fix missing boilerplate in tests

* re-add requirements file to tests folder

* re-enable tests in ci build configuration

* Folder module tests and fixes (#38)

* folder tests wip

* modules/folders: tests and tweaks

* update folders and compute-vm README files

* modules/gcs: tests and minor tweaks

* Create README.md

* Update README.md

* Update README.md

* Update README.md

* Added docker image for strongSwan

* Add support for routes and tests to net-vpc module (#39)

* modules/net-vpc: add routes (untested)

* initial tests

* modules/net-vpc: add test for flow logs

* modules/net-vpc: split tests into two separate files

* modules/net-vpc: routes test

* modules/net-vpc: test routes

* Add support for Terraform plugin cache in ci test build file (#40)

* add Terraform plugin caching to test ci build configuration

* fix mkdir in test build configuration

* trigger test check

* Refactor dynamic vpn configuration for on-prem-in-a-box module

* Fix dynamic vpn for onprem-in-a-box module

* Migrate Shared VPC example to local modules (#41)

* wip

* wip

* validated, untested

* modules/compute-vm: make service account email in locals resilient to destroy

* modules/project: make project id output depend on iam roles

* fixes

* shared-vpc tweaks

* update diagram

* update README input output tables

* modules/compute-vm: add service account IAM email output

* move GKE service account roles at the project level, add GCE service account roles

* update diagram and README

* modules/project: add extra output for IAM-dependent project id

* update modules READMEs

* minor tweaks

* modules/compute-vm: fix service account output

* remove static address from NAT

* fix container service agent binding dependency

* rename shared vpc

* Update README.md

* Update README.md

* Add static vpn gw to on-prem-in-a-box module

* Refactor hub and spoke to use new modules (#42)

* modules/compute-vm: saner defaults for service account scopes

* hub and spoke refactor, docs still missing

* complete hub and spoke

* Update README.md

* Add toolbox docker container, fix gw routing to the internet

* Add DNS Hybrid connectivity parameters

* Fix onprem dns zone for the static vpn configuration

* Added readme.md for on-prem module

* Add new line at the end of the files

* Add boilerplate for cloudbuild config files

* fix boilerplate in strongswan shell script

* Update README.md

* include missing file to fix merge conflict

* remove missing file to fix merge conflict

* include missing file to fix merge conflict (again)

* remove content from spurious file used to avoid merge conflicts

* Add net-vpc-peering module

* Initial commit for hub-and-spoke-peering infrastructure example

* Fix typos in infrastructure/ READMEs

* remove stale file

* use larger resolution version of hub and spoke diagram

* Update README.md

* Update hub-and-spoke-peerings example to use internal modules

* Add initial project tests (#46)

* modules/project: make prefix optional

* initial project module tests

* modules/project: use null for unset parent

* modules/dns: backport PR6 from the CFT dns module

* Add testing resources including on-prem-in-a-box to hub-and-spoke-peerings example

* Fix firewall rules to allow connectivity, switch to custom route advertisement for onprem -> spokes connectivity

* Move locals out of main.tf

* remove ssh tag from compute-vm variable default

* Add ssh tag to the test vms

* Update README.md

* Update README.md

* Update README.md

* Hub and spoke peering changes (#48)

* rename hub-and-spoke-vpn

* add ssh tag to shared-vpc-gke instance

* rename and rework hub and spoke peering

* fix test requirements

* align hub and spoke peering with module contents

* diagram

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* minimal fixes to onprem examples variable files

* onprem example stub, missing DNS zones and private.googleapis records onprem

* add missing boilerplate

* Update README.md

* Update README.md

* infra/onprem: add test instance and minimal outputs

* add DNS modules and resource

* infra/onprem: diagram and initial README

* minor changes to onprem module and example (#49)

* update toolbox image

* infra/onprem: add zone for private access, add metadata domain to onprem dns

* infra/onprem: onnprem service account, add testing procedure in README

* Update README.md

* infra/onprem: remove extra variable

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* infra/onprem: rename forwarder address variable

* Update README:

Added explicit --tunnel-through-iap for gcloud compute ssh commands

* Update top-level and section READMEs (#50)

* top-level README WIP

* rewrite top-level README

* change top-level README title

* remove initial quote in top-level README

* Update README.md

* Update README.md

* Update README.md

* foundations README

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* add experimental scheduled cloud function module

* scheduled cloud function module: allow disabling schedule

* business-units foundation example (#52)

* Added folder-units module.

* Business units example update (WIP)

* Update all BU modules to internal ones

* Refactoring business-units example, add billing and org IAM handling

* update projects tests for new iam additive naming

* update project README for new iam additive naming

* streamline bu example and module (#53)

Co-authored-by: Ludovico Magnocavallo <ludomagno@google.com>

* align net-vpn-ha interface with the other vpn modules

* update module README files

* Update README.md

* Update README.md

* Create CHANGELOG.md

* Refactor COS module to be generic (#51)

* Create generic COS module and update CoreDNS module to use it

* Update compute-vm-cos README

* Fix COS README

* Update COS example

* Skip boilerplate check for COS file template

* Make COS module more generic and provide preset configurations

* Update COS module documentation

* tfdoc: add support for multiple variables files

* compute-vm: split boot disk in separate variable file for cos module support

* Streamline cos modules (#54)

* tfdoc: fix bug in last commit

* compute-vm: add support for user-data

* compute-vm: restore noncos variable split

* remove compute-vm-cos-coredns

* compute-vm: revert to original state

* cos-container/coredns

* fix variables mess

* cos/coredns fixes

* cos/mysql

* remove stale compute-vm-cos module

* add test instance to cos modules

* tfdoc: add support for multiple output files

* cos: add initial READMEs

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* add test apply fixture

* cos-coredns: tested

* Update README.md

* Fix typo

* cos-coredns: refactor README

* Update README.md

* test yaml validity in cos modules tests

* cos mysql tests

* cos mysql: refactor and test (disk tests missing)

* onprem: fix Coredns

* cos mysql: additional disk working

* cos modules: fix instance disks for no instance

* update some modules READMEs

* update some modules READMEs

* Update README.md

* Update README.md

* add simple tests for foundations/environments

* change default for org id in foundations/environments to avoid errors when none is specified

* fix null/empty organization id in foundations/environments

* fix errors when destroying on empty state in foundations/environments

* fundations/bu: fix errors when destroying with empty state

* modules/gcs: make outputs resilient on destroy with empty state

* modules/folders: make outputs resilient on destroy with empty state

* switch organization_id variable to long form in foundations/bu and modules/folders-unit

* Update README.md

* infra/shared-vpc: remove duplicate tag attribute from bastion

Co-authored-by: Aleksandr Averbukh <averbukh@google.com>
Co-authored-by: Julio Castillo <juliocc@gmail.com>
Co-authored-by: Julio Castillo <jccb@google.com>
2020-04-03 14:06:48 +02:00
..
assets Merge development branch (#44) 2020-04-03 14:06:48 +02:00
README.md Merge development branch (#44) 2020-04-03 14:06:48 +02:00
diagram.png Merge development branch (#44) 2020-04-03 14:06:48 +02:00
main.tf Merge development branch (#44) 2020-04-03 14:06:48 +02:00
outputs.tf Merge development branch (#44) 2020-04-03 14:06:48 +02:00
variables.tf Merge development branch (#44) 2020-04-03 14:06:48 +02:00
versions.tf Merge development branch (#44) 2020-04-03 14:06:48 +02:00

README.md

On-prem DNS and Google Private Access

This example leverages the on prem in a box module to bootstrap an emulated on-premises environment on GCP, then connects it via VPN and sets up BGP and DNS so that several specific features can be tested:

The example has been purposefully kept simple to show how to use and wire the on-prem module, but it lends itself well to experimenting and can be combined with the other infrastructure examples in this repository to test different GCP networking patterns in connection to on-prem. This is the high level diagram:

High-level diagram

Managed resources and services

This sample creates several distinct groups of resources:

  • one VPC
  • one set of firewall rules
  • one Cloud NAT configuration
  • one test instance
  • one service account for the test instance
  • one service account for the onprem instance
  • one dynamic VPN gateway with a single tunnel
  • two DNS zones (private and forwarding) and a DNS inbound policy
  • one emulated on-premises environment in a single GCP instance

Cloud DNS inbound forwarder entry point

The Cloud DNS inbound policy reserves an IP address in the VPC, which is used by the on-prem DNS server to forward queries to Cloud DNS. This address needs of course to be explicitly set in the on-prem DNS configuration (see below for details), but since there's currently no way for Terraform to find the exact address (cf Google provider issue), the following manual workaround needs to be applied.

Find out the forwarder entry point address

Run this gcloud command to (find out the address assigned to the inbound forwarder)[https://cloud.google.com/dns/docs/policies#list-in-entrypoints]:

gcloud compute addresses list -project [your project id]

In the list of addresses, look for the address with purpose DNS_RESOLVER in the subnet to-onprem-default. If its IP address is 10.0.0.2 it matches the default value in the Terraform forwarder_address variable, which means you're all set. If it's different, proceed to the next step.

Update the forwarder address variable and recreate on-prem

If the forwader address does not match the Terraform variable, add the correct value in your terraform.tfvars (or change the default value in variables.tf), then taint the onprem instance and apply to recreate it with the correct value in the DNS configuration:

tf apply
tf taint module.on-prem.google_compute_instance.on_prem_in_a_box
tf apply

CoreDNS configuration for on-premises

The on-prem module uses a CoreDNS container to expose its DNS service, configured with foru distinct blocks:

  • the onprem block serving static records for the onprem.example.com zone that map to each of the on-prem containers
  • the forwarding block for the gcp.example.com zone and for Google Private Access, that map to the IP address of the Cloud DNS inbound policy
  • the google.internal block that exposes to containers a name for the instance metadata address
  • the default block that forwards to Google public DNS resolvers

This is the CoreDNS configuration:

onprem.example.com {
  root /etc/coredns
  hosts onprem.hosts
  log
  errors
}
gcp.example.com googleapis.com {
  forward . ${resolver_address}
  log
  errors
}
google.internal {
  hosts {
    169.254.169.254 metadata.google.internal
  }
}
. {
  forward . 8.8.8.8
  log
  errors
}

Testing

Onprem to cloud

# connect to the onprem instance
gcloud compute ssh onprem

# check that the BGP session works and the advertised routes are set
sudo docker exec -it onprem_bird_1 ip route |grep bird
10.0.0.0/24 via 169.254.1.1 dev vti0  proto bird  src 10.0.16.2
35.199.192.0/19 via 169.254.1.1 dev vti0  proto bird  src 10.0.16.2
199.36.153.4/30 via 169.254.1.1 dev vti0  proto bird  src 10.0.16.2

# get a shell on the toolbox container
sudo docker exec -it onprem_toolbox_1 sh

# test forwarding from CoreDNS via the Cloud DNS inbound policy
dig test-1.gcp.example.com +short
10.0.0.3

# test that Private Access is configured correctly
dig compute.googleapis.com +short
private.googleapis.com.
199.36.153.8
199.36.153.9
199.36.153.10
199.36.153.11

# issue an API call via Private Access
gcloud config set project [your project id]
gcloud compute instances list

Cloud to onprem

# connect to the test instance
gcloud compute ssh test-1

# test forwarding from Cloud DNS to onprem CoreDNS (address may differ)
dig gw.onprem.example.com +short
10.0.16.2

# test a request to the onprem web server
curl www.onprem.example.com -s |grep h1
<h1>On Prem in a Box</h1>

Operational considerations

A single pre-existing project is used in this example to keep variables and complexity to a minimum, in a real world scenarios each spoke would probably use a separate project.

The VPN used to connect to the on-premises environment does not account for HA, upgrading to use HA VPN is reasonably simple by using the relevant module.

Variables

name description type required default
project_id Project id for all resources. string
bgp_asn BGP ASNs. map(number) ...
bgp_interface_ranges BGP interface IP CIDR ranges. map(string) ...
ip_ranges IP CIDR ranges. map(string) ...
region VPC region. string europe-west1
resolver_address GCP DNS resolver address for the inbound policy. string 10.0.0.2
ssh_source_ranges IP CIDR ranges that will be allowed to connect via SSH to the onprem instance. list(string) ["0.0.0.0/0"]

Outputs

name description sensitive
foo None
onprem-instance Onprem instance details.
test-instance Test instance details.