SQL Server - Upgrade

PaaS upgrade workflow

Request

The upgrade of a SQL Server PaaS instance is the responsibility of cegedim.cloud. Requests must be submitted via an ITCare ticket, specifying a preferred time window for the operation.

Please indicate if the upgrade must be performed outside of business hours, to allow for proper RFC planning.

It is strongly recommended to first perform the upgrade in a non-production environment, to validate compatibility of the entire process with the client's technical and application constraints.

Upgrades of SQL Server clusters older than version 2016 are not supported by cegedim.cloud.

This limitation is due to major changes introduced in the SQL Server engine starting with version 2016, particularly regarding AlwaysOn features and the deprecation of legacy components.

Major version upgrades (e.g., from SQL Server 2019 to 2022) require a different approach, namely the full reprovisioning of the environment. In this case, a new PaaS instance must be provisioned with the target version, followed by a manual migration of databases.

This process only covers the application of Cumulative Updates (CU) published by Microsoft. These updates contain security, stability, and performance fixes for SQL Server.

Process

The upgrade process involving Cumulative Updates follows these steps:

1. Snapshot of virtual machines (to enable rollback in case of failure)

2. Full backup of customer databases using Rubrik

3. Installation of SQL Server Cumulative Updates (official Microsoft patches)

4. Server reboot to apply the updates

5. Integrity checks on databases and review of error logs

For AlwaysOn clusters:

• Secondary nodes are patched first

• A manual failover to the secondary node is performed prior to updating the primary

• Controlled reboot of the updated node

• The process is repeated for remaining nodes

Impacts

Applying Cumulative Updates can lead to temporary service interruptions or changes in application behavior.

Points to consider:

• Temporary unavailability of SQL Server during reboot

• Risk of functional regressions due to internal behavior changes in CU

• Possible rollback required if installation fails

Potential risks:

• Failure to rejoin the AlwaysOn cluster

• Data corruption if backups are invalid

• Broken application dependencies due to changes introduced by the CU

Estimated Duration

The duration of an upgrade depends on several factors, including the data volume to be backed up and the deployment mode (standalone vs AlwaysOn).

Estimated durations:

• Standalone instance with low volume: ~1 hour

• AlwaysOn cluster with failover: ~2 to 3 hours

These durations are indicative only and should be adjusted according to the specific constraints of the target environment (I/O, load, network latency, etc.).

Operating System Updates

Windows Server updates are performed exclusively during scheduled "Patch Parties"—maintenance windows during which critical security and stability patches are applied in a coordinated manner, in line with cegedim.cloud's maintenance schedule.

These operations are outside the scope of the SQL Server upgrade. If an OS version change is required, the following steps must be taken:

• Request reprovisioning of a PaaS instance with the new OS version via ITCare

• Perform a manual migration of the databases to the new instance

In-place upgrades of the operating system are not supported.