Upgrading GitHub Enterprise Server

Preparing to upgrade

1. Determine an upgrade strategy and choose a version to upgrade to. For more information, see "Upgrade requirements."

2. Create a fresh backup of your primary instance with the GitHub Enterprise Server Backup Utilities. For more information, see the GitHub Enterprise Server Backup Utilities README.md file.

3. If you are upgrading using an upgrade package, schedule a maintenance window for GitHub Enterprise Server end users. If you are using a hotpatch, maintenance mode is not required.

   Note: The maintenance window depends on the type of upgrade you perform. Upgrades using a hotpatch usually don't require a maintenance window. Sometimes a reboot is required, which you can perform at a later time. Following the versioning scheme of MAJOR.FEATURE.PATCH, patch releases using an upgrade package typically require less than five minutes of downtime. Feature releases that include data migrations take longer depending on storage performance and the amount of data that's migrated. For more information, see "Enabling and scheduling maintenance mode."

Taking a snapshot

A snapshot is a checkpoint of a virtual machine (VM) at a point in time. We highly recommend taking a snapshot before upgrading your virtual machine so that if an upgrade fails, you can revert your VM back to the snapshot. If you're upgrading to a new feature release, you must take a VM snapshot. If you're upgrading to a patch release, you can attach the existing data disk.

There are two types of snapshots:

• VM snapshots save your entire VM state, including user data and configuration data. This snapshot method requires a large amount of disk space and is time consuming.

• Data disk snapshots only save your user data.

  Notes:

  • Some platforms don't allow you to take a snapshot of just your data disk. For these platforms, you'll need to take a snapshot of the entire VM.
  • If your hypervisor does not support full VM snapshots, you should take a snapshot of the root disk and data disk in quick succession.

Upgrading with a hotpatch

You can upgrade GitHub Enterprise Server to the latest patch release using a hotpatch, which does not require a maintenance window and usually does not require a reboot. You can use hotpatching to upgrade to a newer patch release, but not a feature release. For example, you can upgrade from 2.10.1 to 2.10.5 because they are in the same feature series, but not from 2.10.9 to 2.11.0 because they are in a different feature series. Using the Management Console, you can install a hotpatch immediately or schedule it for later installation. You can use the administrative shell to install a hotpatch with the ghe-upgrade utility. For more information, see "Upgrade requirements."

Upgrading a single appliance with a hotpatch

Installing a hotpatch using the Management Console

1. Enable automatic updates. For more information, see "Enabling automatic updates."
2. From an administrative account on GitHub Enterprise Server, click in the upper-right corner of any page.
   
3. In the left sidebar, click Management Console.
   
4. At the top of the Management Console, click Updates.
   
5. When a new hotpatch has been downloaded, use the Install package drop-down menu:
   • To install immediately, select Now:
   • To install later, select a later date.
     
6. Click Install.
   

Installing a hotpatch using the administrative shell

1. SSH into your GitHub Enterprise Server instance.

   $ ssh -p 122 admin@HOSTNAME

2. Browse to the GitHub Enterprise Server Releases page. Next to the release you are upgrading to, click Download, then click the Upgrading tab. Copy the URL for the upgrade hotpackage (.hpkg file).
3. Download the upgrade package to your GitHub Enterprise Server instance using curl :

   admin@HOSTNAME:~$ curl -L -O UPGRADE-PKG-URL

4. Run the ghe-upgrade command using the package file name:

   admin@HOSTNAME:~$ ghe-upgrade GITHUB-UPGRADE.hpkg
   *** verifying upgrade package signature...

5. If a reboot is required for updates for kernel, MySQL, Elasticsearch or other programs, the hotpatch upgrade script notifies you.

Upgrading an appliance that has replica instances using a hotpatch

Appliances configured for high-availability and geo-replication use replica instances in addition to primary instances. To upgrade these appliances, you'll need to upgrade both the primary instance and all replica instances, one at a time.

Upgrading the primary instance

1. Upgrade the primary instance by following the instructions in "Installing a hotpatch using the administrative shell."

Upgrading a replica instance

1. Upgrade the replica instance by following the instructions in "Installing a hotpatch using the administrative shell." If you are using multiple replicas for Geo-replication, you must repeat this procedure to upgrade each replica one at a time.

2. Connect to the replica instance over SSH as the "admin" user on port 122:

   $ ssh -p 122 admin@replica-host

3. Verify the upgrade by running:

   $ ghe-version

Upgrading with an upgrade package

While you can use a hotpatch to upgrade to the latest patch release within a feature series, you must use an upgrade package to upgrade to a newer feature release. For example to upgrade from 2.11.10 to 2.12.4 you must use an upgrade package since these are in different feature series. For more information, see "Upgrade requirements."

Upgrading a single appliance with an upgrade package

1. SSH into your GitHub Enterprise Server instance.

   $ ssh -p 122 admin@HOSTNAME

2. Browse to the GitHub Enterprise Server Releases page. Next to the release you are upgrading to, click Download, then click the Upgrading tab. Select the appropriate platform and copy the URL for the upgrade package (.pkg file).

3. Download the upgrade package to your GitHub Enterprise Server instance using curl :

   admin@HOSTNAME:~$ curl -L -O UPGRADE-PKG-URL

4. Enable maintenance mode and wait for all active processes to complete on the GitHub Enterprise Server instance. For more information, see "Enabling and scheduling maintenance mode."

   Note: When upgrading the primary appliance in a High Availability configuration, the appliance should already be in maintenance mode if you are following the instructions in "Upgrading the primary instance."

5. Run the ghe-upgrade command using the package file name:

   admin@HOSTNAME:~$ ghe-upgrade GITHUB-UPGRADE.pkg
   *** verifying upgrade package signature...

6. Confirm that you'd like to continue with the upgrade and restart after the package signature verifies. The new root filesystem writes to the secondary partition and the instance automatically restarts in maintenance mode:

   *** applying update...
   This package will upgrade your installation to version version-number
   Current root partition: /dev/xvda1 [version-number]
   Target root partition:  /dev/xvda2
   Proceed with installation? [y/N]

7. For single appliance upgrades, disable maintenance mode so users can use your GitHub Enterprise Server instance.

   Note: When upgrading appliances in a High Availability configuration you should remain in maintenance mode until you have upgraded all of the replicas and replication is current. For more information, see "Upgrading a replica instance."

Upgrading an appliance that has replica instances using an upgrade package

Appliances configured for high-availability and geo-replication use replica instances in addition to primary instances. To upgrade these appliances, you'll need to upgrade both the primary instance and all replica instances, one at a time.

Upgrading the primary instance

1. On the primary instance, enable maintenance mode and wait for all active processes to complete. For more information, see "Enabling maintenance mode."
2. Connect to the replica instance over SSH as the "admin" user on port 122:

   $ ssh -p 122 admin@replica-host

3. On the replica instance, or on all replica instances if you're running multiple replica instances as part of geo-replication, run ghe-repl-stop to stop replication.
4. Upgrade the primary instance by following the instructions in "Upgrading a single appliance with an upgrade package."

Upgrading a replica instance

1. Upgrade the replica instance by following the instructions in "Upgrading a single appliance with an upgrade package." If you are using multiple replicas for Geo-replication, you must repeat this procedure to upgrade each replica one at a time.

2. Connect to the replica instance over SSH as the "admin" user on port 122:

   $ ssh -p 122 admin@replica-host

3. Verify the upgrade by running:

   $ ghe-version

4. On the replica instance, to start replication, run ghe-repl-start.

5. On the replica instance, to make sure replication services are running correctly, run ghe-repl-status. This command will return OK for all services when a successful replication is in progress and the replica has upgraded. If the command returns Replication is not running, the replication may still be starting. Wait about one minute before running ghe-repl-status again.

   Note: While the resync is in progress ghe-repl-status may return expected messages indicating that replication is behind. For example: CRITICAL: git replication is behind the primary by more than 1007 repositories and/or gists

   If ghe-repl-status didn't return OK, follow the steps below to manually start the replication.

   1. On the replica instance, run ghe-repl-setup <primary-instance-ip> again.
   2. On the replica instance, to start replication, run ghe-repl-start.
   3. On the replica instance, to make sure replication services are running correctly, run ghe-repl-status. This command will return OK for all services when a successful replication is in progress and the replica has upgraded.

6. When you have completed upgrading the last replica, and the resync is complete, disable maintenance mode so users can use your GitHub Enterprise Server instance.

Restoring from a failed upgrade

If an upgrade fails or is interrupted, you should revert your instance back to its previous state. The process for completing this depends on the type of upgrade.

Rolling back a patch release

To roll back a patch release, use the ghe-upgrade command with the --allow-patch-rollback switch. When rolling back an upgrade, you must use an upgrade package file with the .pkg extension. Hotpatch package files with the .hpkg extension are not supported.

ghe-upgrade --allow-patch-rollback EARLIER-RELEASE-UPGRADE-PACKAGE.pkg

A reboot is required after running the command. Rolling back does not affect the data partition, as migrations are not run on patch releases.

For more information, see "Command-line utilities."

Rolling back a feature release

To roll back from a feature release, restore from a VM snapshot to ensure that root and data partitions are in a consistent state. For more information, see "Taking a snapshot."