Cluster Upgrades in AWS Elemental Conductor File - AWS Elemental Conductor File

This is version 2.15 of the AWS Elemental Conductor File documentation. This is the latest version. For prior versions, see the Archive section of AWS Elemental Conductor File and AWS Elemental Server Documentation.

Cluster Upgrades in AWS Elemental Conductor File

In a Conductor cluster, upgrade the AWS Elemental Conductor File nodes first, and then upgrade each of the AWS Elemental Server nodes.

As part of the upgrade of the Conductor node, the Conductor software stops working. When this happens, all of the worker nodes go into "headless" mode. The following rules apply until you upgrade all of the worker nodes:

  • You cannot control the worker nodes from the Conductor web interface.

  • The web interfaces of the worker nodes will have limited functionality: you will only be able to stop and start existing jobs.

During headless mode, activity continues on the worker nodes, but you can't control the worker node.

Upgrade rules

The following rules apply when you're upgrading.

  • Your system must be in a working state prior to the upgrade. If it's in a degraded state (such as not accepting jobs or not responding through the web interface), the upgrade will not work.

  • You can upgrade to a version that's a maximum of two major versions above your current version, such as from 2.14.x to 2.15.x. The number of patches between the two versions is irrelevant. To upgrade over a bigger span, you must perform several upgrades, such as from 2.12.2 to 2.14.4, then to 2.15.0.

  • You must stop the elemental_se service on all AWS Elemental Server nodes before upgrading any Conductor File or AWS Elemental Server nodes in the cluster. For instruction, see the procedures in the following sections.

The upgrade commands are the same, but the procedure around upgrades changes depending on your type of cluster. The following sections provide instructions for the different kinds of AWS Elemental Conductor File clusters.

Important

Plan to upgrade during a maintenance window. All activity on the nodes will stop during downgrade.