Graph refresh

[slemrmartin]

Graph refresh collects inventory to InventoryCollection + InventoryObject objects and saves by Persister.

Avoids using nested hashes when collecting data.

Switching Refresher as ManagerRefresh subclass, added collector/parser/persister classes.

Parser components are derived as much as possible from old style refresh.

---

• depends on InventoryCollection definitions for Lenovo manageiq#18063

[slemrmartin]

[slemrmartin]

cc @agrare @Ladas @douglasgabriel

There is one problem with inventories physical_disks and canisters. They didn't have unique key in old refresh and as I see they didn't have even unique key in lenovo REST API doc.
For canisters, there should be called /canisters to get UUID.
I used serial_number instead, because it sounds like (yes :) unique value

[agrare]

@slemrmartin yeah this is being added by ManageIQ/manageiq-schema#285
If you could review that PR and all of the others that would be a big help

[slemrmartin]

cc @EsdrasVP

[EsdrasVP]

@slemrmartin As I commented on ManageIQ/manageiq-schema#285 (comment) there are cases in which the serial_number isn't returned by the API we consume, I don't surely know if this happens with Canister, but it happens with the drivers we parse to Physical Disks. But if we do what you suggested on the discussion, we can eliminate this from Canister side.

[slemrmartin]

@EsdrasVP okay, I changed it in ManageIQ/manageiq#18063, that unique value is storage id + serial number, which can be nil.
Disk/canister will be badly overwritten only in case of 2 disks/canisters without serial number belonging to the same storage.

Also there are more changes in #235, which must be integrated to this code, right? There are some disk "total space" and multi-enclosures, right?

[EsdrasVP]

Disk/canister will be badly overwritten only in case of 2 disks/canisters without serial number belonging to the same storage.

@slemrmartin Yeah, if what I did on #235 it's approved and we use GET /canisters, this problem should not happen.

Also there are more changes in #235, which must be integrated to this code, right? There are some disk "total space" and multi-enclosures, right?

Yes, although the implementation of total_space was made on another, and already approved, PR. Don't know why is there, maybe due to a rebase, I'll check that.

[slemrmartin]

@EsdrasVP all parser components are copied to new location here, because there are too many changes to use the same code for current and graph refresh.
So during the time there will be 2 parallel refreshes, code must be written to both places.

[slemrmartin]

• integrated Fix the firmware name #232 to graph refresh

[slemrmartin]

#233 integrated

[miq-bot]

Checked commits slemrmartin/manageiq-providers-lenovo@9fc1ab6~...d47706f with ruby 2.3.3, rubocop 0.52.1, haml-lint 0.20.0, and yamllint 1.10.0
28 files checked, 3 offenses detected

app/models/manageiq/providers/lenovo/inventory/collector.rb

• ⚠️ - Line 26, Col 5 - Lint/UselessAssignment - Useless assignment to variable - version_parser.
• ❗ - Line 21, Col 7 - Naming/AccessorMethodName - Do not prefix reader method names with get_.

app/models/manageiq/providers/lenovo/physical_infra_manager/hardware.rb

• ❗ - Line 3, Col 5 - Rails/InverseOf - Specify an :inverse_of option.

[CharlleDaniel]

@slemrmartin I will close and reopen only for the travis to perform the specs again.

[slemrmartin]

@CharlleDaniel this PR depends on ManageIQ/manageiq#18063, so tests are failing now

[Ladas]

Awesome work @slemrmartin 👍