| name | documentdb-azure-deployment |
| description | Deploy an Azure DocumentDB cluster (`Microsoft.DocumentDB/mongoClusters`) end-to-end — Bicep (primary), Azure CLI one-shot, Terraform, or portal. Covers resource-group creation, cluster parameters (tier, storage, server version, sharding, HA), firewall rule configuration, retrieving the connection string, and teardown. Use when the user asks to provision, create, deploy, or spin up an Azure DocumentDB cluster, or wants infrastructure-as-code for one. |
| license | MIT |
Deploy Azure DocumentDB
Interactive skill for provisioning a managed Azure DocumentDB cluster (resource type Microsoft.DocumentDB/mongoClusters, API 2025-09-01). Azure DocumentDB is the managed Azure service built on the open-source microsoft/documentdb engine.
For running DocumentDB locally instead (Docker / Compose), use documentdb-local-deployment. For connection-string tuning after the cluster exists, use documentdb-connection.
No-agent shortcut. A ready-to-run Bicep + deploy script is checked in at examples/azure-deployment/ — customers who prefer to run the deploy themselves can clone the repo and just run ./deploy.sh (no arguments). The script interactively lists subscriptions → resource groups → regions (same flow as this skill). Use it as a reference when generating project files too.
Step 0 — preflight checks (run before anything else)
Run these checks at the start of every deployment session. If any fails, fix it in place and re-run the loop before continuing to Step 1 — don't skip ahead, and don't ask the user configuration questions you'll have to re-ask after they fix their environment.
az version --query '"azure-cli"' -o tsv
az account show --query '{name:name, id:id}' -o json
az account show --query name -o tsv
az provider show --namespace Microsoft.DocumentDB --query registrationState -o tsv
az role assignment list --assignee "$(az ad signed-in-user show --query id -o tsv)" \
--scope "/subscriptions/$(az account show --query id -o tsv)" \
--query "[].roleDefinitionName" -o tsv
az provider show --namespace Microsoft.DocumentDB \
--query "resourceTypes[?resourceType=='mongoClusters'].locations[]" -o tsv
examples/azure-deployment/deploy.sh and deploy.ps1 implement checks 1–4 automatically and will az login / register the provider / create the resource group as needed. If you're generating deployment files into a user's project, copy those scripts or use them as a template.
Step 0.5 — is this production or dev/test?
Ask the user before anything else in Step 1. The answer drives every default. Never assume.
| If "production" (default) | If "dev / prototype / test" |
|---|
| Tier: M30 (minimum that supports HA) | Tier: M10 or M20 |
| Storage: 128 GiB per shard | Storage: 32 GiB |
| HA: ZoneRedundantPreferred | HA: Disabled |
Firewall: Private Endpoint; allowAzureServices: false | allowAzureServices: true + developer IP rule |
| Password source: Key Vault reference | Key Vault preferred; literal OK for throwaway |
Parameters file: main.parameters.sample.json | main.parameters.dev.json |
Production is the safer default — the Bicep template and main.parameters.sample.json in examples/azure-deployment/ ship with M30 + ZoneRedundantPreferred + 128 GiB so that a customer who runs ./deploy.sh <rg> <location> without overrides ends up with a cluster they can actually put workloads on. If the user answers "dev", either:
Step 1 — pick the Azure subscription (always ask, never assume)
The currently active subscription from az account show may not be the one the user wants. Always list and confirm — never silently use the active one.
az account list --query "[].{Name:name, SubscriptionId:id, State:state, IsDefault:isDefault}" \
--output table
Present the list to the user and ask them to pick one (by name or ID). Then set it active so subsequent commands are scoped to it:
az account set --subscription "<subscription-id-or-name>"
az account show --query "{name:name, id:id}" -o table
Record the chosen subscription ID as $SUBSCRIPTION_ID and pass --subscription "$SUBSCRIPTION_ID" to every subsequent az command in this flow — this guarantees RG / region lookups stay scoped to the user's choice even if the active context drifts.
If the user has only one subscription, still confirm out loud ("I'll deploy into <name> — OK?") rather than silently proceeding.
Step 2 — pick the resource group in that subscription (existing or new)
List the resource groups scoped to the chosen subscription (do not omit --subscription — it prevents showing RGs from a different context):
az group list \
--subscription "$SUBSCRIPTION_ID" \
--query "[].{Name:name, Location:location}" \
--output table
Present the list to the user and ask them to pick one of:
(a) Reuse an existing RG — take its location from the table above and use that as the cluster's region by default. Skip Step 3 (the region is already fixed by the RG). Record it as $LOCATION.
(b) Create a new RG — ask the user for the new RG name, then continue to Step 3 to pick a region before creating it.
Do not proceed to Step 3 or Step 4 until the user has explicitly picked (a) or (b).
Step 3 — pick the Azure region (only when creating a new RG)
Reached only when the user chose 2(b). List the regions that support Microsoft.DocumentDB/mongoClusters so the user can pick one that is both regionally appropriate and supported:
az provider show \
--subscription "$SUBSCRIPTION_ID" \
--namespace Microsoft.DocumentDB \
--query "resourceTypes[?resourceType=='mongoClusters'].locations[]" \
--output tsv
Present the list and ask the user to pick one. Record it as $LOCATION. Then create the RG in the chosen subscription + region:
az group create \
--subscription "$SUBSCRIPTION_ID" \
--name "<new-rg-name>" \
--location "$LOCATION"
Confirm creation succeeded before moving on:
az group show --subscription "$SUBSCRIPTION_ID" --name "<new-rg-name>" --query "{name:name, location:location, state:properties.provisioningState}" -o table
Step 4 — gather the remaining cluster inputs
Now that subscription, RG, and location are fixed, ask for cluster-specific values:
| Input | Example | Notes |
|---|
| Cluster name | docdb-prod-001 | 8–40 chars, lowercase letters/digits/hyphens; globally unique in Azure |
| Admin username | clusteradmin | Avoid reserved names like admin, root |
| Admin password | — | 8–128 chars; store in Key Vault — never commit |
| Compute tier | M30 (prod default) / M10–M20 (dev) | Full list: M10, M20, M30, M40, M50, M60, M80, M200 |
| Storage per shard (GiB) | 128 (prod default) / 32 (dev) | |
| Shard count | 1 (sufficient up to TB scale) | |
| High availability | ZoneRedundantPreferred (prod default) / SameZone / Disabled (dev) | Non-Disabled values require M30+ |
| MongoDB server version | 8.0 | |
| Public network access | Default Enabled with firewall rules | Or disable and attach Private Endpoint (see documentdb-security) |
If the user didn't answer Step 0.5, ask again — without that answer you can't pick the right tier/HA defaults.
Step 5 — choose a deployment path
| Path | Use when | Section |
|---|
| Bicep (recommended) | Repeatable infra-as-code, PR-reviewed, committed to repo | Step 6a |
| Azure CLI one-shot | Prototype, local dev, quick validation | Step 6b |
| Terraform | Existing Terraform estate | Step 6c |
| Portal | First-time users who want to see the UI | Azure portal quickstart |
For Bicep, load references/bicep-cluster-template.md — it contains the canonical parameterized template and an optional private-endpoint variant.
Step 6a — deploy with Bicep
Generate main.bicep using the template in references/bicep-cluster-template.md, then:
az login
az account set --subscription "<subscription-name-or-id>"
az group create \
--name "<resource-group-name>" \
--location "<location>"
az deployment group create \
--resource-group "<resource-group-name>" \
--template-file main.bicep
az deployment group create \
--resource-group "<resource-group-name>" \
--template-file main.bicep \
--parameters @main.parameters.json
Secret handling. Never hardcode adminPassword in main.parameters.json or a repo. Options:
Step 6b — deploy with Azure CLI one-shot
For quick iteration without a Bicep file:
az login
az account set --subscription "<subscription-name-or-id>"
az group create --name rg-docdb-dev --location eastus2
az resource create \
--resource-group rg-docdb-dev \
--name docdb-dev-001 \
--resource-type "Microsoft.DocumentDB/mongoClusters" \
--api-version 2025-09-01 \
--location eastus2 \
--properties '{
"administrator": { "userName": "clusteradmin", "password": "REPLACE_WITH_STRONG_PASSWORD" },
"serverVersion": "8.0",
"sharding": { "shardCount": 1 },
"storage": { "sizeGb": 32 },
"highAvailability": { "targetMode": "Disabled" },
"compute": { "tier": "M10" }
}'
az resource create \
--resource-group rg-docdb-dev \
--name "docdb-dev-001/AllowAllAzureServices" \
--resource-type "Microsoft.DocumentDB/mongoClusters/firewallRules" \
--api-version 2025-09-01 \
--properties '{ "startIpAddress": "0.0.0.0", "endIpAddress": "0.0.0.0" }'
Never paste a real password on the command line in shared terminals — read it from an env var or Key Vault.
Step 6c — deploy with Terraform
Prefer this when the user already uses Terraform. The AzureRM provider 4.x exposes azurerm_mongo_cluster, which targets Microsoft.DocumentDB/mongoClusters API 2025-09-01.
Load references/terraform-cluster-template.md for the canonical main.tf (with variables, validation, sample terraform.tfvars, and the firewall-rule pattern via the AzAPI provider). Then:
export ARM_SUBSCRIPTION_ID=$(az account show --query id -o tsv)
export TF_VAR_admin_password=$(az keyvault secret show \
--vault-name <kv-name> --name docdb-admin-password --query value -o tsv)
terraform init -upgrade
terraform plan -out main.tfplan
terraform apply main.tfplan
The provider's high_availability_mode accepts only Disabled and ZoneRedundantPreferred (the API also accepts SameZone, but that mode is not exposed through azurerm_mongo_cluster 4.x — use the Bicep template if you need it). Full Microsoft quickstart: https://learn.microsoft.com/azure/documentdb/quickstart-terraform.
Step 7 — verify the deployment
az resource list \
--resource-group "<resource-group-name>" \
--namespace Microsoft.DocumentDB \
--resource-type mongoClusters \
--query "[].name" \
--output json
Expect one entry matching your cluster name.
Step 8 — retrieve the connection string
From the portal: cluster → Connection strings. The returned string has a <password> placeholder you must substitute.
Form of the connection string:
mongodb+srv://<user>:<password>@<cluster>.global.mongocluster.cosmos.azure.com/?tls=true&authMechanism=SCRAM-SHA-256&retrywrites=false&maxIdleTimeMS=120000
Note retrywrites=false — Azure DocumentDB does not support retryable writes; leaving it at the driver default will cause connection errors (see documentdb-connection for driver-specific tuning).
Step 9 — configure access
Pick one posture and help the user apply it:
-
Public + firewall (dev only). Add the developer's IP:
MY_IP=$(curl -s https://api.ipify.org)
az resource create \
--resource-group rg-docdb-dev \
--name "docdb-dev-001/dev-$(whoami)" \
--resource-type "Microsoft.DocumentDB/mongoClusters/firewallRules" \
--api-version 2025-09-01 \
--properties "{ \"startIpAddress\": \"$MY_IP\", \"endIpAddress\": \"$MY_IP\" }"
The 0.0.0.0–0.0.0.0 rule is the documented shortcut for "Allow Azure services and resources within Azure" — use for serverless workloads. Never leave 0.0.0.0–255.255.255.255 in place outside a short connection test.
-
Private Endpoint (prod). See documentdb-security — public access should be disabled and a Private DNS zone added.
-
Entra RBAC / CMK / diagnostic settings. See documentdb-security and documentdb-monitoring.
Step 10 — teardown
az group delete --name "<resource-group-name>" --yes --no-wait
Confirm with the user before running — this removes everything in the resource group, not just the cluster.
References
Related skills
documentdb-local-deployment — Docker / Compose for running DocumentDB locally
documentdb-connection — connection-string tuning after the cluster exists
documentdb-security — Private Endpoint, Entra RBAC, CMK
documentdb-high-availability — HA, cross-region replica, SLA tiers
documentdb-monitoring — diagnostic settings, slow-query logs