Sequential Management

Sequential Management of Horizontally Scaled Containers

Horizontally scaled layers contain several containers of the same type. The platform can manage these nodes together or process them one by one to reduce the impact of maintenance and deployment operations on service availability.

From the layer controls, all containers can be configured, inspected for logs and statistics, restarted, or redeployed as one logical group. For operations that temporarily interrupt a container, sequential management helps keep at least one node available.

On this page

Operations Supporting Sequential Management

The sequential option is available for the following operations on horizontally scaled nodes:

Container RestartRestarts the nodes one after another instead of restarting the complete layer at once.
Container RedeployRedeploys each container consecutively while other nodes remain operational.
Application DeploymentDeploys an application from an archive or VCS repository across the nodes sequentially.
Cloudlet Limit AdjustmentApplies resource-limit changes to the nodes one by one.

Connection Behaviour

Connections to a node are terminated while an intrusive operation is performed on that node. Sequential processing reduces the overall service impact because the remaining nodes continue operating.

Stateless protocols: HTTP requests can be redirected automatically to another active node, which helps minimise visible downtime.
Stateful protocols: WebSocket, FTP, and SSH connections are not automatically restored. The application must include its own reconnect mechanism; otherwise, active connections are lost even when sequential management is used.

Simultaneous vs Sequential Management

The sequential approach is optional. Choose the management mode according to the environment and operational requirement.

Management ModeHow It WorksAdvantagesRecommended Use
SimultaneousApplies the operation to all containers in one run.Completes the task faster.Development and testing, where a short service interruption is acceptable.
SequentialProcesses the containers one by one with a configured interval.Reduces downtime and keeps other nodes available.Production workloads where service continuity is more important than operation speed.

Sequential Management Parameters

1Delay

Defines the waiting interval between operations on consecutive nodes. The default is 30 seconds, and the maximum is 5 minutes.

2Temporarily Remove Node(s) from DNS

Temporarily excludes the affected node from internal and external DNS records while the operation is running, so new requests are sent to active nodes.

Usage Recommendations

  • Use simultaneous management for faster work in development and testing.
  • Use sequential management for production services that must remain available.
  • Set the delay according to the real startup and health-check time of the application.
  • Enable temporary DNS removal for high-traffic services.
  • Implement reconnection logic for WebSocket, FTP, SSH, and other stateful connections.
  • Keep enough healthy nodes active to handle traffic while another node is unavailable.
Capacity planning: During sequential maintenance, the remaining active nodes must have enough resources to serve the workload without the node being updated.

What’s next?