Disaster recovery procedures, backup and recovery, rollback procedures, and operational guidance for AccessPoint for Microsoft 365, and how recovery works in Hosted AccessPoint
Last updated: November 10, 2026 by Steve
Disaster Recovery
AccessPoint for Microsoft 365 runs entirely in your Azure subscription. Disaster recovery planning leverages Azure's built-in redundancy and recovery capabilities. The publisher does not hold any customer data, so recovery is fully within your control. If your organization uses Hosted AccessPoint, Realizer runs the service and its backups: see Hosted AccessPoint.
Recovery Objectives
These are typical figures for your own deployment. They follow from the Azure choices you make (tiers, redundancy, backup retention, a second region), not from a commitment by Realizer.
| Component | RPO | RTO | Mechanism |
|---|---|---|---|
| Azure SQL Database | Minutes (PITR) | 1–2 hours | Point-in-time restore to a new database |
| Azure Blob Storage | For an overwrite or delete: none lost within 30 days. For a regional loss: depends on the copy you keep | Minutes to hours | Soft delete and versioning; a geo-redundant account or an independent copy you add (the template deploys locally redundant storage) |
| App Service (API) | Zero (stateless) | Minutes | Redeploy from artifacts or swap deployment slots |
| SPFx Web Part | Zero (stateless) | Minutes | Re-upload .sppkg from published artifacts |
| Configuration | Same as database | Same as database | Restored with the database |
Backup and Recovery
Database
- Point-in-time restore (PITR): Automatic. A new installation gets the S0 size, and the template leaves Azure SQL's default of 7 days of point-in-time restore. Standard, Premium and General Purpose sizes allow up to 35 days (Basic at most 7). Raise it under Azure Portal > SQL Database > Backups > Retention policies, or with
az sql db str-policy set -g <rg> -s <server> -n <database> --retention-days 35. Later deployments do not change the value you set. - Long-term retention (LTR): Configure weekly, monthly, or yearly backups for compliance. Set up via Azure Portal > SQL Database > Backups > Retention policies.
- Geo-redundant backup: Azure SQL keeps backups in geo-redundant storage by default, and the template does not set the backup storage redundancy, so the platform default applies. This allows a restore to a paired Azure region in case of a regional outage. Confirm the setting on your database (SQL Database > Backups), and that the paired region meets your data-residency requirements.
Blob Storage
The template configures blob protection, so nothing needs enabling.
- Soft delete: deleted blobs, versions and containers are kept for 30 days and can be restored (open the container, choose Show deleted blobs, then Undelete, or use
az storage blob undelete). - Versioning: every overwrite or delete keeps the previous version. A lifecycle rule deletes previous versions 30 days after they were created (never in the
evidencecontainer), so a bad overwrite or delete stays restorable for at least 30 days. - Purged records: a retention purge or case delete also deletes the previous versions of every file it removes, so the content is permanently gone about 30 days later, when soft delete expires. If you deployed with
grantBlobVersionDeletion=false, the purge log showsVersionsPendingand the lifecycle rule removes those versions within about 60 days. - Do not disable versioning or remove the lifecycle rule, and never add a lifecycle rule to the
evidencecontainer. - To restore a version, list it with
az storage blob list --account-name <storage-account> --container-name documents --prefix <path> --include v --auth-mode loginand copy it over the current blob. - Redundancy: the template creates the documents storage account as locally redundant storage (LRS) and states that on every deployment, so a redundancy you change in the Azure portal may be put back by a later upgrade — check it after each upgrade. For a copy that survives the loss of a region, prefer an independent one: object replication to a second storage account, or a vaulted Azure Backup for blobs, in a region that meets your residency rules.
- Case archiving and redundancy: Azure cannot convert an account from LRS to geo-redundant storage once a file has been archived on LRS, and the Archive tier is not available on zone-redundant storage. Settle redundancy before you switch on a storage policy that archives originals.
Configuration
Export or document your jurisdiction packs before making major changes. If a configuration change needs to be reverted, the only option is to restore from a database backup.
Rollback Procedures
API Rollback
The API is deployed as a zip package to Azure App Service. To revert to a previous deployment:
- Open the Azure Portal and navigate to your App Service
- Go to Deployment Center > Deployment history
- Locate the previous successful deployment
- Select it and click Redeploy
- Verify the application is functioning correctly
If you use deployment slots, swap back to the previous slot via Deployment slots > Swap.
Database Rollback
DacPac deployments are additive by default — new columns and tables are added but existing objects are not removed. This provides a safety net for most schema changes going forward. It does not make an older DacPac safe: a table the DacPac defines is brought to its definition, so an older DacPac would drop the columns, indexes and constraints added since.
That is why an API rollback never rolls the database back. From 2.5.0 the database records the version that last updated it, and an API of 2.5.0 or later that is older than that version leaves the newer schema exactly as it is and runs against it: /api/health/migration reports Degraded with the outcome NewerDatabase, and what the newer version added is simply not used. To apply the older version's schema on purpose — it removes what the newer version added, with its data — set DatabaseMigration__AllowSchemaDowngrade and DatabaseMigration__AllowDataLoss to true for one start, then remove both.
Going back below 2.5.0. A version older than 2.5.0 does not read that record, so its schema update would try to remove everything added since; with data loss blocked (the default) the update stops and that version's API answers 503. Before you redeploy:
- While 2.5.0 or later still runs, choose Remove all training data in Lab mode settings.
- Restore any files a storage policy moved to the Archive tier that you will need; an older version cannot bring them back. Files in Cool and Cold stay readable.
- Add the App Service setting
DatabaseMigration__Enabled=false, so the older version runs against the newer schema without changing it. - Redeploy the older version. With the public portal deployed, the portal web app goes back to the same version.
- Set
DatabaseMigration__Enabledback totruewhen you return to 2.5.0 or later.
For data issues (corrupted or incorrect data), use Azure SQL point-in-time restore:
- Restore the database to a moment before the issue occurred
- Verify the restored database contents
- Update the App Service connection string to point to the restored database, or rename the databases to swap them
SPFx Rollback
To revert the SPFx web part to a previous version:
- Download the previous version's
.sppkgpackage from the Realizer portal - Open the SharePoint App Catalog
- Upload the previous package, replacing the current version
- Click Deploy
Configuration Rollback
Jurisdiction packs are additive — importing a pack adds or updates items but does not remove existing ones. To revert configuration changes, restore the database from a point-in-time backup taken before the change.
Realizer serves each AccessPoint version the jurisdiction packs in the form that version can import, so an API you roll back to an earlier release still receives packs it can read: content added for a later release is left out, and a few items are served as they were released.
Disaster Recovery Scenarios
Scenario 1: Accidental Data Deletion
- Documents deleted from Blob Storage: Recover the blob or an earlier version within the 30-day soft delete and version window
- Database records deleted: Use Azure SQL point-in-time restore to the moment before the deletion. Copy the needed records from the restored database to production.
Scenario 2: Azure Region Outage
If the Azure region hosting your deployment becomes unavailable:
- Azure SQL: Restore from geo-redundant backups to a new SQL Server in a different region
- Blob Storage: If using geo-redundant storage (GRS or RA-GRS), initiate an account failover. If using locally redundant storage (LRS), data must be restored from an external backup.
- App Service: Redeploy using the published template in the recovery region. A recovery resource group counts as a new installation from the template's point of view, so choose New installation and pass the same optional-component parameters you used originally. The template stages the API package itself and the API applies the database schema on startup. Then point the recovered App Service at the restored database and storage account, run the Finish Setup page again for the new managed identity, and set any operator application settings you had customized.
- API URL: Usually no action needed — once the API redeploys and re-validates its license, it self-registers its new address and the web part rediscovers it automatically. Only if you pinned the URL with an
AccessPoint_ApiUrlstorage entity override do you need to update it to the new App Service URL (or repoint your custom domain).
Scenario 3: Full Environment Rebuild
The entire AccessPoint environment can be rebuilt from scratch using the published deployment artifacts:
- Run the published template (New installation, with the same optional-component parameters as before) to provision all Azure resources: App Service, SQL, storage, Key Vault, Application Insights and Log Analytics, the document worker and any optional components. The template stages the API package, and the API creates the schema on first start.
- Reinstall the SharePoint package (from the install page in the deployment's outputs, AppSource or a manual upload) and re-approve its API permissions. Upload the Teams app if it was removed.
- Open the Finish Setup page (
finishSetupUrl) or runDeploy-AccessPoint.ps1 -SkipInfrastructureto grant Graph permissions to the new managed identity, then stop and start the App Service. - Restore the database from the most recent backup (point-in-time or geo-restore), then repoint the App Service at it or swap names.
- Restore the Blob Storage documents from geo-redundant or external backups into the
documentscontainer. - Re-add any operator application settings you had customized.
Infrastructure as Code (Bicep templates) and the deployment script ensure the environment is fully reproducible.
Planning Recommendations
- Keep a copy of your documents outside the region — preferably an independent copy (object replication to a second account, or a vaulted Azure Backup for blobs); if you switch the account to geo-redundant storage instead, check after each upgrade that it still is, and do it before any storage policy archives originals
- Enable long-term retention (LTR) on Azure SQL for weekly, monthly, and yearly backups beyond the default 7-day point-in-time restore window (up to 35 days)
- Keep the template's blob protection in place — 30-day soft delete and blob versioning are configured for you
- Keep the keys with the data. The Key Vault's
dataprotectionkey is the only way to read stored connector and payment secrets and the portal mail key, and a database or storage account under a customer-managed key cannot be read without that key, including a restored database, whose recovery server needs access to every key version its backup used. Keep the key backups, and keep the original vault: a rebuild into a new resource group creates a new vault, but the restored key ring still names the originaldataprotectionkey, so grant the new API identity Key Vault Crypto Service Encryption User on that key. - Document your deployment parameters — keep a record of your tenant name, resource group, alert email, and any custom App Service settings so you can redeploy quickly
- Test recovery annually — restore the database and Blob Storage to a test environment to verify your backups are working
- Consider Azure Front Door or Traffic Manager if your organization requires automatic failover with minimal downtime
- Plan for a lapsed subscription. From 2.6.0 an expired or revoked license makes AccessPoint read-only rather than unavailable, and the API keeps working for 30 days on its last validated license if Realizer cannot be reached. Renew before the grace period ends, and plan how statutory deadlines and requestor contact are kept if the deployment is ever read-only
Hosted AccessPoint
If your organization uses Hosted AccessPoint, Realizer operates the service and its recovery; there is nothing on this page for you to run.
- Where the backups are. Hosted data is stored in Microsoft Azure's Canada Central region. The database backups and the stored files are geo-replicated to Canada East, so they stay in Canada and a restore in Canada East is possible.
- What backups are for. Realizer keeps them to recover the service after an incident. They are not a way to restore records your organization deleted; deleted rows age out of the database's point-in-time backups.
- No promised recovery time. The service is provided without a promised uptime, recovery time or amount of recent work that could be lost in a recovery, unless a written agreement with Realizer says otherwise (see the Hosted AccessPoint terms). Service events are posted on the status page.
- Releases. Realizer releases each version by a warm swap into staging slots and swaps back on any failure, so a failed release returns the service to the previous build without signing anybody out.
- Your own copy. An Administrator can download all of the organization's data at any time from Settings > Subscription — every record and file, requestor details included (Lab mode's fictional cases are left out) — up to three times a month, each kept for 14 days. Case exports and response packages stay available too. Take a copy whenever your own records rules need one.
Health Monitoring
AccessPoint exposes two health check endpoints:
| Endpoint | Purpose |
|---|---|
GET /api/health |
Basic health check — confirms the API process is running |
GET /api/health/ready |
Readiness check — verifies database connectivity, blob storage, and dependent services |
Key Metrics to Watch
| Metric | Where to Find | Warning Threshold |
|---|---|---|
| HTTP 5xx rate | Application Insights > Failures | Any sustained 5xx errors |
| Average response time | Application Insights > Performance | > 3 seconds |
| SQL DTU usage | Azure Portal > SQL Database > Monitoring | > 80% sustained |
| Blob storage capacity | Azure Portal > Storage Account > Metrics | Approaching quota |
| App Service CPU/memory | Azure Portal > App Service > Metrics | > 80% sustained |
Scaling
Vertical Scaling (Scale Up)
Change the App Service plan tier in Azure Portal > App Service > Scale up (App Service plan). No redeployment required.
Horizontal Scaling (Scale Out)
The AccessPoint API is stateless and supports multiple instances. Configure manual or automatic scaling in Azure Portal > App Service > Scale out (App Service plan).
SQL Scaling
Adjust the database tier based on workload. Common service objectives: Basic, S0, S1, S2, S3, P1, P2.
Certificate and Secret Rotation
| Item | Rotation Required | Notes |
|---|---|---|
| Managed identity | No | Credentials managed automatically by Azure |
| Entra ID client secret | Yes, if used | Rotate before expiry; managed identity is preferred |
| SSL/TLS certificate | Depends | App Service Managed Certificate renews automatically; custom certificates must be renewed manually |