Hitachi Block Storage

Hitachi Block Storage

[Cyrus]

Hello!,

Hope you're doing well. I'm scratching a long standing itch in the way
traditional Storage systems are used with PVE, but I would like to
make a stop and check and ask around in the devel makelist.

I'm building a storage plugin to consume storage services from a
Hitachi VSP E-series storage system via fiberchannel. I would like to
achieve vSphere VVol like features and in that order of ideas I was
wondering if a framework should be built instead:

Option 1: any vendor could clone my plugin and adapt it to make it
work with their brand/model of machine (no framework, just forking)
Option 2: Build a vSphere VVol-like framework/plugin, which can accept
different connectors implementing the specific API calls required per
operation.

Option 2 seems the more vendor-friendly way to go, but after reading
https://pve.proxmox.com/wiki/Storage_Plugin_Development it seems it
would be a framework within the framework (odd at least).

I would like to read you feedback, I'm currently halfway to this scope
with Option 1:

* 1 LUN per virtual disk — direct array volumes, no LVM layer.
* Thin provisioning via Hitachi Dynamic Provisioning (DP) pools.
* Snapshots — array-offloaded Thin Image, per LDEV, with metadata
tracked in a cluster-replicated registry.
* Copy-on-write linked clones — space-efficient Thin Image clones from
a base image or a snapshot; full copies are handled by Proxmox via the
device path.
* Online volume resize — array expand + host-side multipath resize.
* QoS — per-LDEV upper/lower IOPS and throughput limits and I/O priority.
* Multipath-aware — FC WWN discovery, ALUA device stanza, automatic
WWID whitelisting (find_multipaths strict), and authoritative WWID
from the array.
* Active-node-only LUN mapping — keeps per-host LUN counts low; live
migration remaps on the fly.
* Management-plane controller redundancy — mgmt_ip accepts multiple
per-controller endpoints with automatic failover and
re-authentication.
* Storage migration — Move Storage to/from file stores (hot/cold),
plus volume_export/volume_import for offline cross-node / pvesm
migration.
* Disk reassignment (rename_volume), base/template images, orphan
detection, and partial-failure rollback during provisioning.
* Replication CLI (hitachiblock-repl) for TrueCopy, Universal
Replicator, and Global-Active Device (GAD).

Also, I don't see any hooks to expose in the web UI the plugin
options, am I missing something?

PS: I'm a happy Ceph user, but I believe this should be written :)

Regards.