Host Connectivity and Masking Views
This domain tests how a LUN reaches a host, why a host sees only some of its paths, and which object you change when connectivity is wrong.
The three building blocks
- Initiator group (IG): one or more host initiators of the same protocol (FC WWN, iSCSI IQN, NVMe NQN). An initiator can belong to only one initiator group, so a failed add usually means it already exists elsewhere.
- Port group (PG): a set of front-end array ports for the same protocol, ideally spread across directors and engines so path loss is survivable.
- Storage group (SG): the devices to be presented.
Masking view rules
- A masking view binds exactly one IG, one PG and one (possibly parent/nested) SG; you can add further IGs or SGs to an existing view, for example
symaccess -sid 123 add view -name APP_MV -ig HOST2_IG. - Zoning must cover every port in the port group to every host HBA. Single-initiator, multiple-target zoning is the normal practice; a host zoned to only one array port has one path and no redundancy.
- LUN IDs are assigned automatically. When a device is removed and re-presented, its LUN ID can change and break host device names, so use consistent LUN assignment where host stability matters.
- iSCSI and NVMe/TCP add their own objects: IP interfaces on ports, target portal groups, CHAP for iSCSI, and separate initiator group types per protocol.
- To withdraw storage, remove the SG (or device) from the view rather than deleting the view, unless the whole presentation is being retired.
Worked example
A Linux host with two HBAs shows four paths to a device, but after adding a second HBA to the initiator group it still shows four. Reasoning order: (1) symaccess -sid 123 show view APP_MV to confirm the new initiator is in the IG; (2) confirm the PG contains the ports the new HBA is zoned to; (3) confirm fabric zoning for the new WWN actually includes those array ports; (4) check host multipath. Here the IG was updated but zoning for the new WWN was missing, so the extra paths never formed. Neither the array port group nor the storage group needed changing.
Traps
- Adding an initiator that already belongs to another IG.
- Zoning to only a subset of the port group, which silently reduces redundancy.
- Trying to present the same device through two storage groups.
- Building an FC initiator group for an NVMe host.
- Assuming LUN numbers stay stable after devices are re-presented.