A clean PowerFlex cluster shutdown is a planned storage outage, not simply a matter of powering off the nodes. Before shutdown, stop every workload that uses PowerFlex volumes, confirm that application I/O has ceased, check the system for active rebuild or rebalance work, and inactivate each Protection Domain. Wait for every domain to report Inactive before powering off the hosts. After startup, verify the required PowerFlex components and system health before reactivating the domains and restoring access to applications.
This walkthrough is based on a PowerFlex 4.5.2 lab and Dell’s documented clean shutdown and startup workflow. Inactivating a Protection Domain temporarily disables application I/O and access to its volumes. The exact host shutdown and startup steps depend on the deployment model; integrated PowerFlex Manager nodes and systems with remote replication may require additional procedures and ordering.
This guide covers a planned shutdown and restart of the entire PowerFlex storage cluster. If you only need to shut down or reboot one SDS node while keeping the rest of the cluster online, follow Safely Reboot a PowerFlex Node: MDM and SDS Maintenance Steps instead; the single-node workflow uses MDM ownership transfer and SDS maintenance mode, not a full-cluster shutdown.
Check the clients/SDC activity
1- Access the MDM primary node by SSH and apply the following command. With that, we can see all connected SDCs and if they are performing any IOPS:
scli --login --management_system_ip pfmp.lab.local --username admin
watch -n1 scli --query_all_sdcAs we can see in the following picture, we have five SDCs, some of which are performing some writes IOPS:
2- Additionally, we need to check the system’s health, filtering by “IOPS”:
watch -n1 "scli --query_all | grep 'IOPS'"
Deactivating the Protection Domain (PD)
3- The next step is deactivating a Protection Domain (PD). It will cause all IOPS to stop (we will have downtime).
Since we have two protection domains, we need to disable each one:
scli --query_all | grep -i "protection"
scli --inactivate_protection_domain --protection_domain_name PD01
scli --inactivate_protection_domain --protection_domain_name PD02
While a protection domain is inactivated, the following activities can take place behind the scenes:
- Determine if any current rebuild/rebalance activities are taking place. If so, the shutdown will be delayed (unless it is forced) until they are finished;
- Block future rebuild/rebalance activities;
- Temporarily disable application I/O and disable access to volumes;
- Move the DRL mode of all SDSs to harden in preparation for restarting the server;
- Reload all SDSs before re-enabling data access.
To inactivate a protection domain through the PowerFlex Manager UI, follow the actions below:
1. On the menu bar, click Block –> Protection Domains;
2. In the list of protection domains, select the relevant protection domain and click More Actions –> Inactivate;
3. In the Inactivate Protection Domain, enter the user password and click Inactivate;
4. Verify that the operation has finished successfully, and click Dismiss.
4- Validate that the Protection Domain has been deactivated (the expected output is “Inactive”):
scli --query_protection_domain --protection_domain_name PD01 | grep –i “Oper”
scli --query_protection_domain --protection_domain_name PD02 | grep –i “Oper”
Shut Down the Required Nodes
5- Now, we can shut down all PowerFlex nodes.
Since we are using a Remote Desktop Manager, we can apply the “shutdown –h now” command on all nodes simultaneously. If you are not using a Remote Desktop Manager or something related to it, access each node and execute the shutdown!
Steps to Power On the Nodes
6- To perform a cluster power-on, first, we need to power on each PowerFlex node.
Verify the system’s health and check if all SDSs and SDCs are up and connected:
scli --query_all_sds
scli –query_all_sdc | grep –i “connected”
7- Activate the Protection Domain (PD). In our example, since we are two Protection Domains, we need to activate both:
scli --activate_protection_domain --protection_domain_name PD01
scli --activate_protection_domain --protection_domain_name PD02
Final Thoughts
The cluster is ready to return to service only after every intended Protection Domain is active, the required SDSs and SDCs are connected, and the system reports a healthy state. For replicated environments, also confirm that the Replication Consistency Group has returned to its expected working state.
Once these checks pass, restore application and VM access in a controlled sequence and verify that workloads can read and write their volumes. Treat the procedure as complete only after both the PowerFlex system and its dependent services have been validated.
External References
- Dell — How to Perform a Clean Shut Down and Power On PowerFlex shutdown guidance covering workload shutdown, client and I/O checks, Protection Domain inactivation, host power-off, component verification, and reactivation.
- Dell — Inactivate a Protection Domain PowerFlex 4.5.x administration procedure for graceful shutdown, including the effects on application I/O, rebuild/rebalance activity, and SDS data handling.
-
Dell — Activate a Protection Domain (SCLI Reference)
Documents the PowerFlex 4.5.x activation command and explains that
--force_activateignores current system errors. - Dell — Power Off the Co-Resident System Additional PowerFlex 4.5.x steps for deployments where management-platform services share the system, including workload shutdown and management database considerations.
- Dell — Restart the Replication Cluster PowerFlex 4.5.x guidance for restarting a replication cluster, including activating the target Protection Domain before the source and validating the RCG afterward.
