Media Gateway Control Protocol (MGCP) Lockstep State Reporting Mechanism
RFC 3992
Network Working Group B. Foster
Request for Comments: 3992 F. Andreasen
Category: Informational Cisco Systems
February 2005
Media Gateway Control Protocol (MGCP)
Lockstep State Reporting Mechanism
Status of This Memo
This memo provides information for the Internet community. It does
not specify an Internet standard of any kind. Distribution of this
memo is unlimited.
Copyright Notice
Copyright (C) The Internet Society (2005).
IESG Note
This document is being published for the information of the
community. It describes a non-IETF protocol that is currently being
deployed in a number of products. Implementers should be aware of
RFC 3015, which was developed in the IETF Megaco Working Group and
the ITU-T SG16, and which is considered the standards-based
(including reviewed security considerations) way to meet the needs
that MGCP was designed to address by the IETF and the ITU-T.
Abstract
A Media Gateway Control Protocol (MGCP) endpoint that has encountered
an adverse failure condition (such as being involved in a transient
call when a Call Agent failover occurred) could be left in a lockstep
state whereby events are quarantined but not notified. The MGCP
package described in this document provides a mechanism for reporting
these situations so that the new Call Agent can take the necessary
fault recovery procedures.
1. Introduction
In the Media Gateway Control Protocol (MGCP) [2], when an endpoint
operating in "step" mode generates a Notify, it will enter the
notification state, where it waits for a response to the Notify.
Furthermore, the endpoint must wait for a new NotificationRequest
before it can resume event processing. As long as the endpoint is
waiting for this NotificationRequest, we say that it is in the
lockstep state.
Foster & Andreasen Informational [Page 1]
RFC 3992 MGCP Lockstep State Reporting Mechanism February 2005
An endpoint that is in the lockstep state cannot perform any event
processing and therefore also cannot generate a new Notify.
Endpoints should only be in the lockstep state for a very short time.
However, in adverse conditions, an endpoint could potentially end in
the lockstep state without the Call Agent realizing it. Clearly,
this could have very negative consequences in terms of the service
provided.
The Lockstep package defined in this document defines extensions to
the EndpointConfiguration and RestartInProgress commands that allow a
Call Agent to request an endpoint to inform it when the endpoint is
in the lockstep state for a specified period of time.
1.1. Conventions Used in This Document
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in BCP 14, RFC 2119 [1].
2. Lockstep Package
Package Name: LCK
Version: 0
Package Description: The purpose of this package is to provide a
mechanism for reporting a condition in which an endpoint has been in
the "lockstep state" for a specified period of time.
There are two aspects of this package:
* The ability for a Call Agent to request endpoints to report if
they are in lockstep state for a specified period of time.
This is done with the EndpointConfiguration command, as
described in section 2.1.
* The reporting mechanism itself, which is done with a new
"lockstep" RestartMethod for the RSIP command as described in
section 2.2.
2.1. Request to Report Lockstep State
The new "LCK/LST" EndpointConfiguration parameter is used by the Call
Agent to request the reporting of "lockstep" state. It uses the
following ABNF:
"LCK/LST:" 0*WSP LSTIME
LSTIME = 1*(4DIGIT)
Foster & Andreasen Informational [Page 2]
RFC 3992 MGCP Lockstep State Reporting Mechanism February 2005
where LSTIME is expressed in seconds, with a value ranging from 0 to
9999. A value greater than 2*T-HIST (refer to [2]) is RECOMMENDED.
LSTIME is the amount of time the endpoint is in the lockstep state
before reporting. The timer starts when the endpoint enters the
lockstep state and is canceled if the endpoint leaves the lockstep
state before the timeout occurs. The value provided remains in
effect until explicitly changed (or a restart occurs). If the
endpoint is already in the lockstep state when a non-zero timer value
is provided, the lockstep timer is simply started with the value
provided; any existing lockstep timer is cancelled. The value zero
is used to turn off reporting.
Show full document text