· May 24 15m read


If you're running IRIS in a mirrored configuration for HA in GCP, the question of providing a Mirror VIP (Virtual IP) becomes relevant. Virtual IP offers a way for downstream systems to interact with IRIS using one IP address. Even when a failover happens, downstream systems can reconnect to the same IP address and continue working.

The main issue, when deploying to GCP, is that an IRIS VIP has a requirement of IRIS being essentially a network admin, per the docs.

To get HA, IRIS mirror members must be deployed to different availability zones in one subnet (which is possible in GCP as subnets always span the entire region). One of the solutions might be load balancers, but they, of course, cost extra, and you need to administrate them.

In this article, I would like to provide a way to configure a Mirror VIP without using Load Balancers suggested in most other GCP reference architectures.



We have a subnet running across the region (I simplify here - of course, you'll probably have public subnets, arbiter in another az, and so on, but this is an absolute minimum enough to demonstrate this approach). Subnet's CIRD is, which means it is allocated IPs to As GCP reserves the first and last two addresses, we can use to

We will implement both public and private VIPs at the same time. If you want, you can implement only the private VIP.


Virtual Machines in GCP have Network Interfaces. These Network Interfaces have Alias IP Ranges which are private IP addresses. Public IP Addresses can be added by specifying Access Config

Network Interfaces configuration is a combination of Public and/or Private IPs, and it's routed automatically to the Virtual Machine associated with the Network interface. So there is no need to update the routes. What we'll do is, during a mirror failover event, delete the VIP IP configuration from the old primary and create it for a new primary. All operations to do that take 5-20 seconds for Private VIP only, from 5 seconds and up to a minute for a Public/Private VIP IP combination.

Implementing VIP

  1. Allocate IP address to use as a public VIP. Skip this step if you want private VIP only.
  2. Decide on a private VIP value. I will use
  3. Provision your IRIS Instances with a service account

- compute.instances.get
- compute.addresses.use
- compute.addresses.useInternal
- compute.instances.updateNetworkInterface
- compute.subnetworks.use

For External VIP you'll also need:
- compute.instances.addAccessConfig
- compute.instances.deleteAccessConfig
- compute.networks.useExternalIp
- compute.subnetworks.useExternalIp
- compute.addresses.list

  1. When a current mirror member becomes primary, we'll use a ZMIRROR callback to delete a VIP IP configuration on another mirror member's network interface and create a VIP IP configuration pointing at itself.

That's it.


NotifyBecomePrimary() PUBLIC {
    #include %occMessages
    set sc = ##class(%SYS.System).WriteToConsoleLog("Setting Alias IP instead of Mirror VIP"_$random(100))
    set sc = ##class(%SYS.Python).Import("set_alias_ip")
    quit sc

And here's which must be placed into mgr\python directory:

This script adds Alias IP ( to the VM Network Interface.

You can allocate alias IP ranges from the primary subnet range, or you can add a secondary range to the subnet
and allocate alias IP ranges from the secondary range.
For simplicity, we use the primary subnet range.

Using google cli, gcloud, this action could be performed in this way:
$ gcloud compute instances network-interfaces update <instance_name> --zone=<subnet_zone> --aliases=""

Note that the command for alias removal looks similar - just provide an empty `aliases`:
$ gcloud compute instances network-interfaces update <instance_name> --zone=<subnet_zone> --aliases=""

We leverage Google Compute Engine Metadata API to retrieve <instance_name> as well as <subnet_zone>.

Also note

Google Cloud uses the first two and last two IPv4 addresses in each subnet primary IPv4 address range to host the subnet.
Google Cloud lets you use all addresses in secondary IPv4 ranges, i.e.:
- - Network address
- - Default gateway address
- - Second-to-last address. Reserved for potential future use
- - Broadcast address

After adding Alias IP, you can check its existence using 'ip' utility:
$ ip route ls table local type local dev eth0 scope host proto 66

import subprocess
import requests
import re
import time
from import compute_v1

METADATA_HEADERS = {"Metadata-Flavor": "Google"}
project_path = "project/project-id"
instance_path = "instance/name"
zone_path = "instance/zone"
network_interface = "nic0"
mirror_public_ip_name = "isc-mirror"
access_config_name = "isc-mirror"
mirror_instances = ["isc-primary-001", "isc-backup-001"]

def get_metadata(path: str) -> str:
    return requests.get(METADATA_URL + path, headers=METADATA_HEADERS).text

def get_zone() -> str:
    return get_metadata(zone_path).split('/')[3]

client = compute_v1.InstancesClient()
project = get_metadata(project_path)
availability_zone = get_zone()

def get_ip_address_by_name():
    ip_address = ""
    client = compute_v1.AddressesClient()
    request = compute_v1.ListAddressesRequest(
        filter="name=" + mirror_public_ip_name,
    response = client.list(request=request)
    for item in response:
        ip_address = item.address
    return ip_address

def get_zone_by_instance_name(instance_name: str) -> str:
    request = compute_v1.AggregatedListInstancesRequest()
    request.project = project
    instance_zone = ""
    for zone, response in client.aggregated_list(request=request):
        if response.instances:
            if"{availability_zone}*", zone):
                for instance in response.instances:
                    if == instance_name:
                        return zone.split('/')[1]
    return instance_zone

def update_network_interface(action: str, instance_name: str, zone: str) -> None:
    if action == "create":
        alias_ip_range = compute_v1.AliasIpRange(
    nic = compute_v1.NetworkInterface(
        alias_ip_ranges=[] if action == "delete" else [alias_ip_range],
    request = compute_v1.UpdateNetworkInterfaceInstanceRequest(
    response = client.update_network_interface(request=request)
    print(instance_name + ": " + str(response.status))

def get_remote_instance_name() -> str:
    local_instance = get_metadata(instance_path)
    return ''.join(mirror_instances)

def delete_remote_access_config(remote_instance: str) -> None:
    request = compute_v1.DeleteAccessConfigInstanceRequest(
    response = client.delete_access_config(request=request)

def add_access_config(public_ip_address: str) -> None:
    access_config = compute_v1.AccessConfig(
        name = access_config_name,
    request = compute_v1.AddAccessConfigInstanceRequest(
    response = client.add_access_config(request=request)

# Get another failover member's instance name and zone
remote_instance = get_remote_instance_name()
print(f"Alias IP is going to be deleted at [{remote_instance}]")

# Remove Alias IP from a remote failover member's Network Interface
# TODO: Perform the next steps when an issue will be closed:
# - update google-cloud-compute pip package to a version containing fix (>1.15.0)
# - remove a below line calling gcloud with
# - uncomment update_network_interface() function[
    "--zone=" + get_zone_by_instance_name(remote_instance),
# update_network_interface("delete",
#                          remote_instance,
#                          get_zone_by_instance_name(remote_instance)

# Add Alias IP to a local failover member's Network Interface

# Handle public IP switching
public_ip_address = get_ip_address_by_name()
if public_ip_address:
    print(f"Public IP [{public_ip_address}] is going to be switched to [{get_metadata(instance_path)}]")


Now let's deploy this IRIS architecture into GCP using Terraform and Ansible. If you already running IRIS in GCP or using a different tool, the ZMIRROR script is available here.


We'll need the following tools. As Ansible is Linux only I highly recommend running it on Linux, althrough I confirmed that it works on Windows in WSL2 too.


$ gcloud version
Google Cloud SDK 459.0.0


$ terraform version
Terraform v1.6.3


$ python3 --version
Python 3.10.12


$ ansible --version
ansible [core 2.12.5]


$ ansible-playbook --version
ansible-playbook [core 2.12.5]


If you're running in WSL2 on Windows, you'll need to restart ssh agent by running:

eval `ssh-agent -s`

Also sometimes (when Windows goes to sleep/hibernate and back) the WSL clock is not synced, you might need to sync it explicitly:

sudo hwclock -s

Headless servers

If you're runnning a headless server, use gcloud auth login --no-browser to authenticate against GCP.


We leverage Terraform and store its state in a Cloud Storage. See details below about how this storage is created.

Define required variables

$ export PROJECT_ID=<project_id>
$ export REGION=<region> # For instance, us-west1
$ export TF_VAR_project_id=${PROJECT_ID}
$ export TF_VAR_region=${REGION}
$ export ROLE_NAME=MyTerraformRole
$ export SA_NAME=isc-mirror

Note: If you'd like to add Public VIP which exposes IRIS Mirror ports publicly (it's not recommended) you could enable it with:

$ export TF_VAR_enable_mirror_public_ip=true

Prepare Artifact Registry

It's recommended to leverage Google Artifact Registry instead of Container Registry. So let's create registry first:

$ cd <root_repo_dir>/terraform
$ cat ${SA_NAME}.json | docker login -u _json_key --password-stdin https://${REGION}
$ gcloud artifacts repositories create --repository-format=docker --location=${REGION} intersystems

Prepare Docker images

Let's assume that VM instances don't have an access to ISC container repository. But you personally do have and at the same do not want to put your personal credentials on VMs.

In that case you can pull IRIS Docker images from ISC container registry and push them to Google container registry where VMs have an access to:

$ docker login
$ <Put your credentials here>

$ export IRIS_VERSION=2023.

$ cd docker-compose/iris
$ docker build -t ${REGION}${PROJECT_ID}/intersystems/iris:${IRIS_VERSION} .

$ for IMAGE in webgateway arbiter; do \
    docker pull${IMAGE}:${IRIS_VERSION} \
    && docker tag${IMAGE}:${IRIS_VERSION} ${REGION}${PROJECT_ID}/intersystems/${IMAGE}:${IRIS_VERSION} \
    && docker push ${REGION}${PROJECT_ID}/intersystems/${IMAGE}:${IRIS_VERSION}; \

$ docker push ${REGION}${PROJECT_ID}/intersystems/iris:${IRIS_VERSION}

Put IRIS license

Put IRIS license key file, iris.key to <root_repo_dir>/docker-compose/iris/iris.key. Note that a license has to support Mirroring.

Create Terraform Role

This role will be used by Terraform for managing needed GCP resources:

$ cd <root_repo_dir>/terraform/
$ gcloud iam roles create ${ROLE_NAME} --project ${PROJECT_ID} --file=terraform-permissions.yaml

Note: use update for later usage:

$ gcloud iam roles update ${ROLE_NAME} --project ${PROJECT_ID} --file=terraform-permissions.yaml

Create Service Account with Terraform role

$ gcloud iam service-accounts create ${SA_NAME} \
    --description="Terraform Service Account for ISC Mirroring" \
    --display-name="Terraform Service Account for ISC Mirroring"

$ gcloud projects add-iam-policy-binding ${PROJECT_ID} \
    --member="serviceAccount:${SA_NAME}@${PROJECT_ID}" \

Generate Service Account key

Generate Service Account key and store its value in a certain environment variable:

$ gcloud iam service-accounts keys create ${SA_NAME}.json \

$ export GOOGLE_APPLICATION_CREDENTIALS=<absolute_path_to_root_repo_dir>/terraform/${SA_NAME}.json

Generate SSH keypair

Store a private part locally as .ssh/isc_mirror and make it visible for ssh-agent. Put a public part to a file

$ ssh-keygen -b 4096 -C "isc" -f ~/.ssh/isc_mirror
$ ssh-add  ~/.ssh/isc_mirror
$ ssh-add -l # Check if 'isc' key is present
$ cp ~/.ssh/ <root_repo_dir>/terraform/templates/

Create Cloud Storage

Cloud Storage is used for storing Terraform state remotely. You could take a look at Store Terraform state in a Cloud Storage bucket as an example.

Note: created Cloud Storage will have a name like isc-mirror-demo-terraform-<project_id>:

$ cd <root_repo_dir>/terraform-storage/
$ terraform init
$ terraform plan
$ terraform apply

Create resources with Terraform

$ cd <root_repo_dir>/terraform/
$ terraform init -backend-config="bucket=isc-mirror-demo-terraform-${PROJECT_ID}"
$ terraform plan
$ terraform apply

Note 1: Four virtual machines will be created. Only one of them has a public IP address and plays a role of bastion host. This machine is called isc-client-001. You can find a public IP of isc-client-001 instance by running the following command:

$ export ISC_CLIENT_PUBLIC_IP=$(gcloud compute instances describe isc-client-001 --zone=${REGION}-c --format=json | jq -r '.networkInterfaces[].accessConfigs[].natIP')

Note 2: Sometimes Terraform fails with errors like:

Failed to connect to the host via ssh: kex_exchange_identification: Connection closed by remote host...

In that case try to clean a local ~/.ssh/known_hosts file:

$ for IP in ${ISC_CLIENT_PUBLIC_IP} 10.0.0.{3..6}; do ssh-keygen -R "[${IP}]:2180"; done

and then repeat terraform apply.

Quick test

Access to IRIS mirror instances with SSH

All instances, except isc-client-001, are created in a private network to increase a security level. But you can access them using SSH ProxyJump feature. Get the isc-client-001 public IP first:

$ export ISC_CLIENT_PUBLIC_IP=$(gcloud compute instances describe isc-client-001 --zone=${REGION}-c --format=json | jq -r '.networkInterfaces[].accessConfigs[].natIP')

Then connect to, for example, isc-primary-001 with a private SSH key. Note that we use a custom SSH port, 2180:

$ ssh -i ~/.ssh/isc_mirror -p 2180 isc@ -o ProxyJump=isc@${ISC_CLIENT_PUBLIC_IP}:2180

After connection, let's check that Primary mirror member has Alias IP:

[isc@isc-primary-001 ~]$ ip route ls table local type local dev eth0 scope host proto 66

[isc@isc-primary-001 ~]$ ping -c 1
PING ( 56(84) bytes of data.
64 bytes from icmp_seq=1 ttl=64 time=0.049 ms

Access to IRIS mirror instances Management Portals

To open mirror instances Management Portals located in a private network, we leverage SSH Socks Tunneling.

Let's connect to isc-primary-001 instance. Note that a tunnel will live in a background after the next command:

$ ssh -f -N  -i ~/.ssh/isc_mirror -p 2180 isc@ -o ProxyJump=isc@${ISC_CLIENT_PUBLIC_IP}:2180 -L 8080:

Port 8080, instead of a familiar 52773, is used because we start IRIS with a dedicated WebGateway running on port 8080.

After successful connection, open in a browser. You should see a Management Portal. Credentials are typical: _system/SYS.

The same approach works for all instances: primary (, backup ( and arbiter ( Just make an SSH connection to them first.


Let's connect to isc-client-001:

$ ssh -i ~/.ssh/isc_mirror -p 2180 isc@${ISC_CLIENT_PUBLIC_IP}

Check Primary mirror member's Management Portal availability on Alias IP address:

$ curl -s -o /dev/null -w "%{http_code}\n"

Let's connect to isc-primary-001 on another console:

$ ssh -i ~/.ssh/isc_mirror -p 2180 isc@ -o ProxyJump=isc@${ISC_CLIENT_PUBLIC_IP}:2180

And switch the current Primary instance off. Note that IRIS as well as its WebGateway is running in Docker:

[isc@isc-primary-001 ~]$ docker-compose -f /isc-mirror/docker-compose.yml down

Let's check mirror member's Management Portal availability on Alias IP address again from isc-client-001:

[isc@isc-client-001 ~]$ curl -s -o /dev/null -w "%{http_code}\n"

It should work as Alias IP was moved to isc-backup-001 instance:

$ ssh -i ~/.ssh/isc_mirror -p 2180 isc@ -o ProxyJump=isc@${ISC_CLIENT_PUBLIC_IP}:2180
[isc@isc-backup-001 ~]$ ip route ls table local type local dev eth0 scope host proto 66


Remove infrastructure

$ cd <root_repo_dir>/terraform/
$ terraform init -backend-config="bucket=isc-mirror-demo-terraform-${PROJECT_ID}"
$ terraform destroy

Remove Artifact Registry

$ cd <root_repo_dir>/terraform
$ cat ${SA_NAME}.json | docker login -u _json_key --password-stdin https://${REGION}

$ for IMAGE in iris webgateway arbiter; do \
    gcloud artifacts docker images delete ${REGION}${PROJECT_ID}/intersystems/${IMAGE}
$ gcloud artifacts repositories delete intersystems --location=${REGION}

Remove Cloud Storage

Remove Cloud Storage where Terraform stores its state. In our case, it's a isc-mirror-demo-terraform-<project_id>.

Remove Terraform Role

Remove Terraform Role created in Create Terraform Role.


And that's it! We change networking configuration pointing to a current mirror Primary when the NotifyBecomePrimary event happens.

Author would like to thank @Mikhail Khomenko, @Vadim Aniskin, and @Evgeny Shvarov for the Community Ideas Program which made this article possible.

Discussion (1)2
Log in or sign up to continue