Add acctz counters for login/logout records

[nmahabaleshwar]

Add new counters for acctz records generated for login and logout records

Change Scope

• Added source counters which will be incremented when login and logout records are generated.
• A new type called session-service is added which defines new enums LOGIN and LOGOUT.
• service-request has a new enum SESSION_SERVICE
• This change is not backwards compatible.

For implementations which envelope LOGIN/LOGOUT record for every OPERATION record, we need a specific accounting source counters. Login/Logout events are generic and they cannot always be tied to an operation. Hence this PR introduces counters specific source counters for LOGIN and LOGOUT events.

This is also helpful when a LOGIN record is generated for a failed authentication event where we had ambiguity on which source counter to increment. This is now resolved with a specific counter for the LOGIN event.

For implementations which account using ONCE, we already have specific gRPC/Command counters for the intended operation.

Platform Implementations

None

[dplore]

/gcbrun

[OpenConfigBot]

No major YANG version changes in commit ba3f4bf

[haussli]

Did we discuss how/if to account for SESSION_STATUS_ONCE, et al?

I think that the commit message should be more verbose; it should describe the reason for this change, please.

[nmahabaleshwar]

Did we discuss how/if to account for SESSION_STATUS_ONCE, et al?

I think that the commit message should be more verbose; it should describe the reason for this change, please.

Did you mean the counter update for SESSION_STATUS_ONCE ? SESSION_STATUS_ONCE already has information about which operation is being accounted, gRPC vs command services and we already have counters for them.

I will update the commit description with more details on why this change is required.

[dplore]

/gcbrun

[haussli]

Did we discuss how/if to account for SESSION_STATUS_ONCE, et al?
I think that the commit message should be more verbose; it should describe the reason for this change, please.

Did you mean the counter update for SESSION_STATUS_ONCE ? SESSION_STATUS_ONCE already has information about which operation is being accounted, gRPC vs command services and we already have counters for them.

I will update the commit description with more details on why this change is required.

OK, understood; if login succeeds, the service type is known for the following record.