Save/restore vlan reservations

[sajith]

Resolves #215. The API is really simple, and is implemented using @property:

# Get the current VLAN allocation table.
vlan_table = temanager.vlan_tags_table

# Restore VLAN allocation table.
temanager.vlan_tags_table = vlan_table

We do some checks in the setter, such as: the input data is in the expected shape, and that we have no existing allocations. The assumption is that SDX Controller would attempt to restore VLAN allocation state only during startup.

[coveralls]

Pull Request Test Coverage Report for Build 12038444933

Details

• 17 of 17 (100.0%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.4%) to 91.239%

---

Totals
Change from base Build 11916789690:	0.4%
Covered Lines:	1185
Relevant Lines:	1290

---

💛 - Coveralls

[YufengXin]

For its said purpose, it looks nice. But-:) we may need to have
(1) the instructions for sdx-controller to use the two methods? why the sdx-controller needs to restore the vlan table during startup? who creates the table and calls the setter?

(2) the second scenario: sdx-controller works for some time with some connection provisioned (ie, some vlans allocated from some or all domains), then it crashes and needs a restart: right now, I think the logic is to read the topologies from DB, the existing bandwidth and vlan allocation information are lost?

[sajith]

@YufengXin Good catches. Thank you!

(1) the instructions for sdx-controller to use the two methods? why the sdx-controller needs to restore the vlan table during startup? who creates the table and calls the setter?

SDX Controller will have to maintain the table, because state management is better left to the application. As long as PCE is a library, it should not be maintaining any state between process restarts. Currently SDX Controller already restores some PCE state (topologies), and it should do the rest of it too.

(2) the second scenario: sdx-controller works for some time with some connection provisioned (ie, some vlans allocated from some or all domains), then it crashes and needs a restart: right now, I think the logic is to read the topologies from DB, the existing bandwidth and vlan allocation information are lost?

True, we need to restore that also. Thank you for the reminder! Perhaps that could be implemented with a separate issue and PR?

[YufengXin]

1. A PR in SDX-controller for the corresponding state persistence.

2. Another PR for persistent state after crash recovery

[sajith]

Perhaps we could add a new TEManager.restore_state() method, and that would be simpler to use:

temanager.restore_state(
    topologies  = [topology1, topology2, topology3],
    vlan_table = vlan_table,
    bandwidth_allocation = bandwidth_allocation,
    latency_allocation = latency_allocation
)

This could be even simpler if we have a corresponding get_state() that returns all the state we care about, and then SDX Controller would not have to care about all these details:

# At various SDX Controller checkpoints:
state = temanager.get_state()
db.upsert("temanager-state", state)

# During SDX Controller startup:
state = db.read("temanager-state")
temanager.restore_state(state)

Basically SDX Controller would have to call get_state() and save the results in the DB when there are events that cause state to change (a new connection is made, a link was cut, service degraded... etc?), and then call restore_state() at startup.

Just thinking aloud. Perhaps there is a better way to do this.