Home > Storage > PowerScale (Isilon) > Product Documentation > Management and Migration > Dell PowerScale: Non-Disruptive Upgrade Best Practices > Rolling and simultaneous patch upgrade
Before OneFS 8.2.0, in the patch upgrade command isi upgrade patch, there is a parameter -- rolling=true/false which controls how the patch is applied. With a different purpose of the patch, this parameter behaves differently.
When the parameter --rolling=false is set, a simultaneous patch request is made and a patch will be installed simultaneously across all nodes. The patch will typically run an isi services command to disable and then later to re-enable the affected services. Since services that are affected by the patch are simultaneously restarted on all the PowerScale nodes in the cluster, this will affect the specified services across the entire cluster causing temporary service disruption. For a simultaneous patch request, the PowerScale nodes will not be rebooted.
When the parameter --rolling=true is set, a rolling patch upgrade request is made. In this case, the patch will be installed and the node will be rebooted in succession. For rolling patch upgrade request, the specified services will not be restarted. Instead, NDU will migrate all user connections away from the nodes before starting the patch request. This migration process can be disruptive. NFS with dynamic pool and SMB CA can help to make this migration transparent to the client application. For other workloads, they will get disconnected and need to re-establish the connection when the node they are connected to reboots.
Simultaneous patch upgrade request can save time and act more efficient, but the specified service will be restarted which causes a temporary service disruption. At the same time, rolling patch upgrade request is the default setting and can take much longer especially when the PowerScale cluster is large. But, it can be less disruptive with the combination of NFS dynamic pool or SMB CA.
The guideline is for the specified service which will not impact the workload like WebUI, PAPI, and so on, use the parameter --rolling=false for the patch upgrade. This will make the upgrade more efficient and will not impact the real workload. For a service that can impact the workload, such as NFS or SMB, use the parameter --rolling=true with the combination of NFS dynamic IP pool or SMB CA to minimize the impact to the client application.
When the parameter --rolling=false is set, a simultaneous patch request is made and a patch will be installed simultaneously across all nodes. And in this case, it will reboot all nodes in the cluster simultaneously.
When the parameter --rolling=true is set, a rolling patch upgrade request is made. In this case, it will install the patch and then reboot each node in succession.
The guideline is to use –rolling=true to have minimal impact on the workload. However, if customers are willing to have a maintenance window with the disconnected workload, use – rolling=false to make this patch upgrade more efficient.
Starting from OneFS 8.2.0, the parameter --rolling is no more and now it uses --simultaneous for the same purpose. It is important to carefully read the Readme file for each patch, which explains the behavior of the patch installation process and its impact in details.