netgroup-polito / frog4-orchestrator Goto Github PK

When the frog receives a graph update, it creates again the inter-domain connections.
For instance, if it selected GRE 0x25 at the time of deploying the graph, when such a graph is updated in selects a different GRE key to implement the same connection.

This behaviour derives from the fact that, in the past, the frog orchestrator didn't have state.
Instead now it stores information on the graphs, then, in case of update, it can reuse the same key previously used.

Can we remove the submodule https://github.com/netgroup-polito/vnf-template-library/tree/29230402eb5fa54f67d0e67112ac469887d4f8ef , since this issue have been closed?

The Frog4-orchestrator still have template API and many functions used for the datastore, So this code should be removed from the Frog4-orchestrator.

Currently the scheduler select feasible domains for a given VNF by checking the functional capabilities of that domain.
But also if a domain is able to implement a VNF, this cannot be feasible due to geagraphical problems (the domain has no connection with the end-points defined in the nffg).
Such a check should be added somewhere.

The FROG4 orchestrator does not use the datastore.
Then that information should be removed from the configuration file https://github.com/netgroup-polito/frog4-orchestrator/blob/master/config/default-config.ini#L42-L53, as well as the README should be updated https://github.com/netgroup-polito/frog4-orchestrator/blob/master/README_INSTALL.md#frog-orchestrator-configuration-file .

Note that also the main README of the FROG project must be updated as well https://github.com/netgroup-polito/frog4/blob/master/docs/README_INSTALL.md#frog4-orchestrator , last bullet.

On https://github.com/netgroup-polito/frog4-orchestrator/blob/domain-capabilities/orchestrator_core/scheduler.py
Multiple calls of the function getInterface:
https://github.com/netgroup-polito/frog4-orchestrator/blob/domain-capabilities/orchestrator_core/scheduler.py#L201
https://github.com/netgroup-polito/frog4-orchestrator/blob/domain-capabilities/orchestrator_core/scheduler.py#L231
https://github.com/netgroup-polito/frog4-orchestrator/blob/domain-capabilities/orchestrator_core/scheduler.py#L398
https://github.com/netgroup-polito/frog4-orchestrator/blob/domain-capabilities/orchestrator_core/scheduler.py#L412

The function was removed from https://github.com/netgroup-polito/frog4-orchestrator/blob/domain-capabilities/orchestrator_core/domain_info.py
on the following commit d8c2855
done by @pamaglio91

If a class specification is modified usually it should be checked on all the code that the removed methods are not used anymore.

There are no API to create or updating the user on the database it should be done manually by SQL and it is not described on the README.

It instead uses the name.

Actually, also every domain orchestrator has this problem.

Validate the incoming NFFG request with the NFFG schema.
In addition, since we may need some optional tags in the request (e.g., domain ID), add the proper code that raises an error when the orchestrator is connected to multiple domains, and the NFFG does not specify which domain refers to.
However, do not raise this exception if the orchestrator is connected to a single domain, which will be useful to allow an initial deployment (e.g., for testing), which typically happens with a single domain and we don't want to bother the tester with useless errors.

When the graph is deployed, the returned id is not always correct.
If the id starts with '0', the first zero is truncated.
For example: the graph id is: 09234, the put of the graph returns 9234.
The bug doesn't appear into the get of all graphs, but only in the put of a graph.

When neither a graph is associated to a domain, nor its VNFs, let's deploy the graph on a "default" domain, which is specified by configuration. The new configuration option has already been added; we need to add the code as well.

Distinguish between new graph and update, using the PUT and POST methods. Currently, we use the PUT for both the operations.

The FROG4 orchestrator should not retrieve the VNF template from the datastore.
It should schedule according to the functional-capabilities.
The template should only be used by the domain orchestrators.

The southbound API of the FROG orchestrator towards the domain orchestrators should be changed.
It must be the same already implemented between the GUI and the FROG orchestrator.