managed-nodes-rolloutThe managed-nodes-rollout stage performs a reboot of the managed compute and application nodes in order to reboot them
to a new image and configuration. The system must be configured to use BOS v2 as
it is used to perform the reboot. The reboot operations use the BOS session templates created during
the prepare-images stage. The -mrs stage argument is only valid for compute nodes, since application nodes are not
controlled
by workload manager software. The -mrs reboot argument will reboot all compute and application nodes immediately if
the --limit-managed-rollout argument is not specified.
managed-nodes-rollout details are explained in the following sections:
The managed-nodes-rollout stage changes the running state of the system. It uses Boot Orchestration Service (BOS) v2 session templates to reboot
the specified compute/application nodes.
The following arguments are most often used with the managed-nodes-rollout stage. See iuf -h and iuf run -h for
additional arguments.
| Input | iuf Argument |
Description |
|---|---|---|
| Activity | -a ACTIVITY |
Activity created for the install or upgrade operations |
| Managed rollout strategy | -mrs {reboot,stage} |
Reboot the managed nodes immediately or stage the new image for the WLM to reboot |
| Limit managed rollout list | --limit-managed-rollout LIMIT_MANAGED_ROLLOUT |
List of managed nodes to be rolled out, specified by xnames or HSM node group |
The code executed by this stage exists within IUF. See the managed-nodes-rollout entry
in /usr/share/doc/csm/workflows/iuf/stages.yaml and the corresponding files
in /usr/share/doc/csm/workflows/iuf/operations/
for details on the commands executed.
The managed-nodes-rollout IUF operation is deemed successful if all the initiated BOS v2 sessions are started and
completed. This operation is only deemed a failure if any of the BOS v2 sessions fail to start. Completion of this
operation does NOT mean that the nodes were able to successfully reboot or be configured via CFS. It simply means the
BOS v2 sessions completed. It is important to carefully read the IUF standard output (stdout) during this operation as a
scenario where the reboot or configuration failed on some nodes is possible.
Output like this appears at the end of the managed-nodes-rollout operation:
INFO Session 9769d735-4037-4500-b008-00067b4822ad: 0% components succeeded, 100% components failed
ERROR cfs configuration failed: {'count': 8, 'list': 'x3000c0s29b2n0,x3000c0s29b4n0,x3000c0s31b2n0,x3000c0s29b3n0,x3000c0s31b4n0,x3000c0s31b3n0,x3000c0s31b1n0,x3000c0s29b1n0'}
From the perspective of managed-nodes-rollout this operation succeeded because the BOS v2 sessions completed, and IUF
will report it as such. However, as can be seen in the output, all eight nodes failed to configure.
Debugging resources: To further debug why the configuration failed on the specified nodes see CFS Sessions.
There are multiple resources to further debug why nodes failed to boot. Each document starts with Troubleshoot.
See Rolling Upgrades Using BOS for details on rebooting managed compute and application nodes with BOS v2.
(ncn-m001#) Execute the managed-nodes-rollout stage for activity admin-230127 using the default stage rollout
strategy and limiting the operation to the HSM node group compute-partition-1.
iuf -a admin-230127 run --limit-managed-rollout compute-partition-1 -r managed-nodes-rollout