All worker NCNs are configured to be potential iSCSI targets.
The selective iSCSI worker node personalization feature allows administrators to specify which worker NCNs are enabled as
iSCSI targets. The mechanism by which an administrator specifies this is the iscsi_worker
component group in the
Hardware State Manager (HSM).
The existence of this group on the system means that selective iSCSI worker node personalization is enabled.
If the group does not exist, then the feature is disabled, and all worker nodes will be enabled as iSCSI targets.
While it is technically possible to create this group and leave it empty, this would mean that no worker nodes would be enabled as iSCSI targets, which in turn would mean that no managed nodes would be able to boot. Therefore CSM treats it as an error if the HSM group exists but contains no worker nodes.
Any members of the group that are not worker nodes are ignored and have no effect on iSCSI enablement.
The Group commands section goes over the commands used to create, remove, or modify the group. The Procedures section discusses the different contexts in which this may be done, and any additional steps that must be performed to effectuate the changes.
This section contains example commands for managing selective worker node personalization for iSCSI SBPS.
For more in-depth information on managing HSM groups in general, see Manage Component Groups.
(ncn-mw#
) Create the iscsi_worker
group.
The
description
field is not required to be specified; it has no operational effect.
cray hsm groups create --label iscsi_worker --description "iscsi node personalization" --members-ids x3000c0s5b0n0,x3001c0s35b0n0
Example output:
[[results]]
URI = "/hsm/v2/groups/iscsi_worker"
The output does not contain the list of members in the group. If an administrator wishes to confirm that the group was created with the correct members, see Listing group members.
(ncn-mw#
) Add an additional worker node to the iscsi_worker
group.
Note: Each command may only add a single worker.
cray hsm groups members create --id x3000c0s18b0n0 iscsi_worker
Example output:
[[results]]
URI = "/hsm/v2/groups/iscsi_worker/members/x3000c0s18b0n0"
(ncn-mw#
) List the members of the iscsi_worker
group.
cray hsm groups members list iscsi_worker
Example output:
ids = [ "x3000c0s5b0n0", "x3001c0s35b0n0", "x3000c0s18b0n0",]
(ncn-mw#
) Remove a worker node from the iscsi_worker
group.
Each command may only remove a single worker.
cray hsm groups members delete x3000c0s18b0n0 iscsi_worker
Example output:
code = 0
message = "deleted 1 entry"
(ncn-mw#
) Delete the iscsi_worker
group.
cray hsm groups delete iscsi_worker
Example output:
code = 0
message = "deleted 1 entry"
The procedures vary for enabling or disabling this feature, or modifying the set of worker nodes that are enabled as iSCSI targets. It depends on whether this is being done during a fresh install of CSM, during an upgrade from CSM 1.6 to CSM 1.7, as part of adding or removing a worker NCN, or as a procedure on an operational CSM 1.7+ system.
During a CSM install, if the administrator wants to enable this feature, they should
create the iscsi_worker
HSM group as part of the
iSCSI SBPS configuration
step in the Configure Administrative Access procedure.
The feature will take effect when Management Node Personalization runs on the NCNs, after SAT Bootprep has run.
If the group is not created at this time, and an administrator wishes to enable this feature post-install, additional manual steps are required. This is also true if an administrator creates the group during the install, but later wants to disable iSCSI on some of the workers where it had previously been enabled.
For details, see After initial CSM 1.7 install or upgrade.
During an upgrade from CSM 1.6 to CSM 1.7, if the administrator does not wish to enable this feature, no special steps
need to be taken. If the administrator does wish to enable this feature, then at the beginning of
2. Execute the IUF management-nodes-rollout
stage,
the administrator must create the iscsi_worker
HSM group.
See Creating the group.
When adding or removing a worker NCN,
additional steps may be required. Before performing the add or remove procedure for the worker NCN,
read the section that applies to the current operation and whether or not the iscsi_sbps
group currently exists:
Notes:
- If unsure whether or not the group exists, try Listing the group members.
- If it succeeds, then the group exists (and its member list will be shown).
- If the group does not exist, then the command will fail with an error that it cannot find the group.
- Both before and after the overall add or remove procedure, administrators may wish to validate the iSCSI configuration. See iSCSI SBPS Verification.
iscsi_sbps
group does not existiscsi_sbps
group existsiscsi_sbps
group does not existiscsi_sbps
group existsiscsi_sbps
group does not existThis means that all current worker NCNs are enabled as iSCSI targets. If the new worker NCN being added should also be enabled as an iSCSI target, then no additional steps are required. Skip the rest of this document and perform the procedure to add the worker NCN to the system.
Otherwise (if the new worker NCN should NOT be configured as an iSCSI target), perform the following steps before following the add NCN procedure:
(ncn-mw#
) Get the xnames of all of the worker NCNs currently in the system.
WORKER_XNAMES=$(cray hsm state components list --role Management --subrole Worker --type Node --format json \
| jq -r .Components[].ID | sort -u | tr '\n' , | sed 's/,$//')
echo ${WORKER_XNAMES}
Example output:
x3000c0s7b0n0,x3000c0s9b0n0,x3000c0s38b0n0,x3000c0s11b0n0
Create the iscsi_sbps
group.
The iscsi_sbps
group needs to be created and contain the xnames of all of the worker NCNs except the one being added
(the same list obtained in the previous step). See Creating the group.
Skip the rest of this document and perform the procedure to add the worker NCN to the system. When the new worker performs Management Node Personalization, it will not be enabled as an iSCSI target.
iscsi_sbps
group existsIf the new worker NCN being added should not be configured as an iSCSI target, then no additional steps are required. Skip the rest of this document and perform the procedure to add the worker NCN to the system.
Otherwise (if the new worker NCN should be configured as an iSCSI target), perform the following steps:
(ncn-mw#
) Get the xnames of all of the worker NCNs currently in the system.
echo $(cray hsm state components list --role Management --subrole Worker --type Node --format json \
| jq -r .Components[].ID | sort -u)
Example output:
x3000c0s11b0n0 x3000c0s38b0n0 x3000c0s7b0n0 x3000c0s9b0n0
Proceed with the add worker NCN procedure until the xname for the new worker NCN exists in HSM.
This can be checked by running the same command as in the previous step, and comparing the output. As soon as a new xname shows up in the list, that is the xname of the worker being added.
Add the xname of the new worker node to the iscsi_sbps
group.
Skip the rest of this document and perform the rest of the procedure to add the worker NCN to the system. When new worker performs Management Node Personalization, it will be enabled as an iSCSI target.
iscsi_sbps
group does not existThis means that all worker NCNs are configured as iSCSI targets. In this scenario, the remove NCN procedure can be performed as usual. After it is done, the administrator must refresh the iSCSI configuration. See Refresh iSCSI configuration.
iscsi_sbps
group existsCheck if the worker being removed is not a member of the iscsi_sbps
group.
See Listing group members for how to see the current members of the group.
If the worker being removed is not a member of the group, then skip the rest of this document and perform the procedure to remove the worker NCN from the system. No additional steps are required in this case. Otherwise, if the worker being removed is a member of the group, then proceed with the remaining steps in this section.
Remove the worker being removed from the group.
Refresh the iSCSI configuration.
Skip the rest of this document and perform the procedure to remove the worker NCN from the system.
IMPORTANT: If this is being done in the context of adding or removing a worker NCN, see Adding or removing worker NCN instead.
During normal system operation, this feature can be enabled or disabled, or the set of worker nodes being enabled for iSCSI can be changed. To do this, use the following procedure:
Create, modify, or delete the iscsi_worker
HSM group.
iscsi_worker
HSM group exists),
then there are two options:
Refresh the iSCSI configuration.
Optional: Verify the iSCSI configuration.
After the CFS node personalization is complete for all of the worker NCNs, administrators may wish to verify the iSCSI configuration. See iSCSI SBPS Verification.
The iSCSI configuration on the system should be refreshed whenever any of the following things takes place:
(ncn-mw#
) The refresh_worker_iscsi_config.py
script is provided to modify the state of the worker NCNs in the
Configuration Framework Service (CFS). It modifies the
state in a way that forces them to re-run the iSCSI configuration layer.
- The latest CSM documentation RPM must be installed on the node where this step is being performed. See Check for latest documentation.
- This script can be run on any master or worker NCN where the documentation RPM is installed.
- This script is run without arguments.
- The script only needs to be run once on the system (not once per worker).
/usr/share/doc/csm/scripts/operations/configuration/refresh_worker_iscsi_config.py
Example output:
The following workers will have their CFS states updated and error counts reset: ['x3000c0s11b0n0', 'x3000c0s26b0n0', 'x3000c0s33b0n0', 'x3000c0s7b0n0', 'x3000c0s9b0n0']
Updating CFS components...
Update successful. CFS batcher should soon start sessions to configure the nodes whose states were updated
SUCCESS
Soon after the script completes, CFS batcher will begin scheduling new
CFS sessions to perform the iSCSI refresh.
This may take up to a few minutes, depending how the the
CFS Options are set on the system.
For more information, see
CFS-Batcher
scheduling.
Once these CFS sessions complete successfully, then the iSCSI configuration has been refreshed.